Moonbug Entertainment, the children’s entertainment studio that makes the hugely popular shows Cocomelon, Little Baby Bum, Blippi, and Mia’s Magic Playground has asked its animators to start using artificial intelligence while making its shows, 404 Media has learned.

Moonbug’s Generative AI policy and its “Studio AI Bible,” a guide to using AI to help generate content, seen by 404 Media, explain in detail how its AI use will work. The policies indicate that the company is in the initial stages of experimenting with AI in the creation of its shows, but that for the moment it has put several guardrails in place in part over legal and copyright concerns associated with using AI in a more substantial way. 

“Keep a human in the loop when using AI. AI is used to assist the artist, not be the artist,” the guidelines say. “Always remember that we can only own (copyright) what a human has created and so we need to ensure the final execution involves substantial human creative input. Our core IP (e.g. key characters/signature worlds and important backgrounds) must reflect substantial human creative input and intentionality.” 

The documents show in detail how a major studio, which runs an empire of content that is wildly popular with infants, toddlers, and their families, is using AI. Moonbug’s shows have hundreds of millions of subscribers on YouTube, and many of its shows have spinoff series and movies that air across several major streaming services. Cocomelon spinoffs air on Netflix, the studio is working on a Cocomelon movie for Universal, and the series is set to move to Disney+ next year. YouTube has been flooded with AI-generated content for infants and toddlers, but these documents indicate that even the biggest companies in the industry are using AI, albeit with far more thought and care than slop purveyors. Its properties also have various popular children and baby toys.

“Like many media companies, we're exploring how AI tools can support our creative and production teams. Today, generative AI is not used in episodes of our content,” a Moonbug spokesperson told 404 Media. “Our core principle is that AI should assist the artist, not be the artist. Our guidelines prohibit AI from originating key creative elements such as new characters, core storylines and song lyrics, and require substantial human creative input. Everything we produce — whether they incorporate AI-assisted elements or not — goes through our human-led creative and quality-control process, including frame-by-frame human eyeball review, and rounds of iterations and notes from our creative and production teams.”

The guidelines say that AI can be used for ideation, scripting, storyboarding, design, and animation, but has put guardrails on how AI can be used in each step of the process. 

For example, the company says animators can use AI for “utility tasks/standard production enhancers,” but cannot use it to “alter a VO/actor’s performance without checking with Legal.” It says AI can be used “to refine a human-authored draft,” but “no ghost-writing. Do not generate key creative elements (character arcs, core plot twists, song lyrics) from scratch by AI. The narrative ‘soul’ and key dialogue beats must remain human-authored.” The company says AI can be used to “generate visual research, vibe boards, brainstorm ideas and concepts and exploring texture/colour/lighting references,” but that there is “no ‘prompt to product.’ […] do not move a 100% AI-generated design directly into the production pipeline. It must be translated into a studio-drawn concept to ensure it meets our technical standards and style.”  

At times, the guide gets very granular. It says workers are allowed to use AI to create “generic” designs and textures, such as trees in a background or furniture, for character outfit changes, or for “creating 3D turnarounds from existing human-created 2D characters,” but usually cannot use it to create wholly new characters. “The [AI] assets must be processed by an artist (eg over-painted, tweaked) to ensure human in loop. All new characters should be human created (unless have had approval for specific IPs). Props — if we are creating something unique and to be heavily featured in the series e.g. the Clubhouse — it must be human created. All franchise specific worlds should be human created (unless intended to be generic).”

The company is requiring employees to “first test all tools with non Moonbug IP and/or test assets before proceeding with legal approval.” After being approved by legal, employees are allowed to use AI to help create assets for Moonbug shows, but all AI use, including the prompts used to generate assets, must be cataloged and saved. The company spokesperson told 404 Media that “non Moonbug IP” refers to generic assets that the company has made: “We created generic test characters and environments so teams can freely explore new tools without using Moonbug IP or anyone else's IP. If a tool proves useful, further testing with Moonbug IP requires appropriate approvals.”

“Making content for young children comes with a particular responsibility, regardless of the tools involved,” the spokesperson added. “Our GenAI guidelines add another layer of guardrails: no ‘prompt-to-product,’ human authorship of core creative elements, approved tools and legal review, protections around performers and third-party IP, and human review of finished work. Ultimately, people make the creative decisions about what is appropriate and worthy of our audience.”

All AI tools that the company uses must have individual legal approval, and all AI-generated assets must be kept in separate file folders than human-created ones using a system it is calling “provenance and isolation,” according to the guidelines. The “goal” of this file management system is “to ensure we always have a path back to human-authored works to maintain copyright ownership over what matters.” 

“Every project must have a dedicated folder for AI generated assets. No file from this folder should be moved into main production folders without being processed by human artist [sic] first. All 100% human created assets must also be stored and labeled properly,” the guidelines say. Workers must log all of their AI prompts, then also write an explanation of how a human transformed the asset if it is ultimately used in production. 

A section of the document called “prompt guidance” says that workers should begin the AI production process by “upload[ing] a human-drawn sketch/Moonbug owned IP as a structural reference, use AI as a refiner eg to add texture/lighting.” It says “No text-to-image prompting (unless for brainstorming/research,” and “no style mimicking or soundalikes — do not use prompts like ‘in the style of Pixar/Ghibli’ etc. Use descriptive artistic terms instead — eg ‘hand-painted water colour aesthetic’ or ‘vibrant street-art aesthetic, chromatic aberration.’”

In May, workers with the International Alliance of Theatrical Stage Employees working on a live action Cocomelon spinoff show went on strike, alleging that they were not being provided fair wages and benefits.

“We see AI as a tool that can expand what talented creative people are able to do, not a substitute for the people who create our stories,” the Moonbug spokesperson said. “We're exploring whether these tools can reduce repetitive work, enhance the creative process and potentially allow our teams to create more stories and experiences for families.”

In July, we wrote about the ‘ChatGPT flyer pandemic,’ in which, as you’ve surely seen, businesses everywhere are advertising themselves and their products with AI-generated posters that all look the same. Surf lesson coaches; guitar stores; bars; so many takeout places. They’re all doing it. And all the posters suck ass.

Naturally, a new meta is emerging, where some businesses and organizations are now holding up crudely drawn signs saying they’ll never use AI, or fuck AI, or various things like this. And people absolutely love it: some of these posts have tens of thousands of likes.

“I would rather your event flyer look like this than see more AI slop,” one flyer posted to Instagram, written in pencil on a page of lined paper, reads. It then shows a couple of stick figures smiling, with the text “I will come to this event.”

Another Instagram post has a couple of people from a bar holding a clip board. The hand-written note in one picture says, “We will NOT be using any AI post to promote the following…” Other pictures then ask people to apply for job vacancies at the bar.

A library posted a very similar message on a piece of cardboard, that reads, “We will not be using AI to create content.”

A cafe posted a series of photos on Instagram of someone holding a paper sign that says, “We will not be using any AI posters to promote ourselves. Instead I will be using this paper and marker pen.” One of the photos adds, “the pancakes are really really GOOD!”

“I hear ya’ll hate AI flyers! So here I am doing hair,” a hand-written sign from a stylist reads, complete with a picture of someone working on another’s hair.

Rachel Karten, writer of the Link In Bio newsletter, posted some other examples 404 Media came across as well. 

This trend will probably go stale fast. But, hey, jump on it while you can.

Last week we published a story that revealed an Amazon warehouse where the company scans and destroys thousands of books for AI training data. The following is an interview with one of the Amazon employees at this warehouse. We granted this employee anonymity because they were not authorized to speak to the press. 

The employee worked at Amazon’s VGT3 warehouse, which is housed in the same facility as LAS8, where Amazon operates its print-on-demand business. Both operations are part of a larger complex of Amazon facilities in Las Vegas, Nevada. 

We discovered that VGT3 was used to destructively scan books for AI training data by placing a tracking device in a shipment of rare books that a bookseller suspected was being acquired by an anonymous AI company. We saw the shipment travel across the country before finally landing at VGT3. Online, Amazon employees who worked at VGT3 described the work of receiving shipments of books, cutting the spines off of them, and scanning the loose pages. 

The Amazon employee I talked to described what that operation looks like from the warehouse floor. 

This interview has been edited for clarity and length. 

404 Media: What do you do when you get to work at VGT3?

Anonymous: We have to walk past everything to get to the back of the warehouse. That's where I first saw all the scanners, and I was really wondering what was going on. I remember I even asked someone what they were doing and they said they’re not sure even the people who work there know what they’re doing. Of course, once I got over there and started putting two and two together, I kind of understood. And you know, I think people who work there probably know what was going on over there, but probably just didn't want to talk about it. 

We also saw them cutting the spines off the books and everything. And after they scan it, they throw all the loose leaf papers into a big shuttle. [Editor’s note: Shuttles is how Amazon refers to gaylords, which are big open cardboard boxes that are often used to store books in bulk.] So they're all mixed together. There's no way to put the book back together or anything. Where I worked, primarily they would bring us boxes of books, or sometimes there's shuttles of books.

What kind of books?

All kinds of books. When we first started a lot of them were brand new. Some of them are used as well. Like you could tell, they were liquidated from a library or something like that. We were even getting boxes of stuff from London. There were a lot of books from the University of London. There was even like, I don't know what you would call them, but it was like kind of stapled together papers that said that they were presented to Parliament on the behalf of Her Royal Majesty the Queen. So they were like government documents. I don't know if they're public. Probably they're public, but it was just interesting that we found all that in there. 

What did you think the books were for?

That's what we were wondering, and we were kind of like, ‘this is a lot of random information.’ They have literally any kind of book you could think of. They have ones in different languages. We saw a lot of German and Russian books, and we actually got whole pallets of Japanese books. Some of these books are brand new and they're like still sealed. A lot of the Japanese ones were, but yeah, we do get the stuff from the libraries, and that's the stuff I think is more mixed because they're not all nice in boxes and everything like that. They're kind of just thrown into the shuttle. 

What was it like working there?

Our job was to unbox the books, and honestly, they seem like a mess over there. Like they're not really well managed or organized at all. Their process wasn't solid. Their process changed every day. But we were pretty much putting them in a tote [a small plastic stackable bin] so that they can go be scanned in and sorted, so that they can go over to the people to cut the bindings off. 

They take them off the truck. The dock's right there, and then they stage them, which is just putting them on standby. People pick up the books and bring them to our station and we stack up the totes until it’s a whole pallet, and then the people with the pallet jacks come and they take the whole pallet.

And I think the next area, I think they call it receiving. What I've heard they do there is that they're scanning all the barcodes pretty much to see if they need it and just weed out duplicates. Something else they had us do was—we call it the trash but we don't we don't know exactly what they do with them—but one day we were taking all the duplicates and all the ones they said, “oh, we don't need these,” and we were throwing those in different shuttles. They're like yeeted in there, like they're not nicely placed in there or anything. And they say they're returning them to the vendors, but I'm like, you know, ‘do the vendors go through it and see which ones are actually worth it?’ Because it's like a tall shuttle that's probably six feet tall, six or seven feet tall. They're just piled in there. It’s also right next to where they start cutting the spines, and so I think right after that, that's where it's going [the trash]. 

What does it look like when they cut the spines off the books?

It's a machine people operate. Each of these machines is just like a little workstation, and there'll be one person at each of these stations. It's really safe because it has a little cover and it has a little area where you slide the book into, remove your fingers from the area, and then you press a button, and it just comes down and slices it. And then you remove the books after the blade is gone. 

After they cut the spines they have these kind of library carts, kind of like how you would stack  books, but instead, it's a stacks of paper and little cardboard things to separate—I'm guessing— the different books.

I saw the pages being thrown in a shuttle after they were scanned. We were walking by and we went over to one of the shuttles and looked into it because we could tell they were throwing the old books in it or the cut books in it, and it's just a bunch of loose paper like just sheets thrown in there.

Did you see the scanners?

They probably have like at least 20 scanners, maybe 25. They look like machines that count cash, it flips through the pages really fast. It kind of has something like that they load it into, and I can see their screens, and on their screens, you can see the scans of the book, and it's going really fast. 

What do you think about Amazon doing this?

At first what they were telling people is that it was for Kindles, but I just didn't believe that. I instantly was like, I don't think that's how they do it because of publication rights and stuff, copyright and everything. Then I saw it was for AI. I don't like it only because I wish I could take these books. There's so much knowledge and so much stuff in them and some of them look like they might be rare, and I've heard that they order rare books, and that's why I say some of these are so obscure. I definitely don't like the idea that they can't be reused or anything like that afterwards. They're reducing the amount of available copies for other people.

Ukrainian drones struck the Afipsky oil refinery in southern Russia overnight on 25 August, monitoring Telegram channels reported. Regional authorities claimed falling drone debris caused fires. Ukraine's General Staff later confirmed the strike, and, in the same post, a hit on a gas processing plant in Astrakhan Oblast the night before. 

Fire at the plant, and Russia's debris account

Initially, the Ukrainian monitoring Telegram channels Exilenova+ and Supernova+ reported the strike, publishing footage of the fire and placing one burning storage tank on the plant's grounds. NASA's FIRMS satellite fire data confirmed the burning site, Militarnyi reported.

The Krasnodar Krai operational headquarters stated that falling drone parts started the fire. Governor Veniamin Kondratyev claimed that debris fell on the grounds of the Afipskaya railway station, killing two people and injuring two more, and stated that fires broke out in the nearby settlement and at the refinery itself.

Ukrainian forces struck two U-272 gas separation units at the Astrakhan gas processing plant the night before, the General Staff wrote in the same post. That plant makes technical sulfur, which goes into explosives. 

Drones also attacked Rostov Oblast overnight, with the Novoshakhtinsk refinery as the target, according to Supernova+.

The refinery Ukraine keeps targeting

Drones have reached the plant repeatedly this year, in January, March, June, and July, making the latest hit the fifth since the start of the year, The Moscow Times counted. In the July case, the Russian news Telegram channel Astra said the fire started near the tank farm.

Ukrainian drones have attacked Russian refineries more than 70 times since the start of 2026 and knocked several dozen plants out of service, The Moscow Times said, while Ukraine's General Staff counted 194 strikes on refineries in the first half of the year alone. Five stopped processing entirely after strikes since 1 August: Permnefteorgsintez, Orsknefteorgsintez, Zapsibneftekhim, and the Saratov and Volgograd refineries, the outlet said.

Mathematicians are raising concerns that the technology could kill their profession. But they still have abilities AI doesn’t

Earlier this month, about 40 top mathematicians gathered at OpenAI’s offices to discuss the future of their profession. The meeting was off-the-record, but if recent articles by mathematicians are any guide, it was mostly pretty glum. People fear for their jobs, their careers and the work they love.

We think the contrary view is more likely, at least in the short-term. AI models are nowhere near as capable as experienced academic mathematicians.

Ukrainian drones worked over dozens of Russian military and industrial sites overnight on 24 August, reaching from occupied Ukraine and the Black Sea to southern Russia, the commander of Ukraine's Unmanned Systems Forces (SBS) reported. The night's targets included a fighter jet at a Russian airbase, air-defense radars, fuel and propellant facilities, and vessels of Russia's sanctions-evading shadow fleet. A fuller breakdown released later in the day named the site of each strike and the unit that carried it out. Footage released with the breakdown shows the strikes.

Ukraine has been running mid-range strikes across the belt from occupied Crimea to Rostov Oblast for months, taking out the radars, missile batteries, and the aircraft that fly air-defense patrols over it — a rolling effort to thin the cover that meets every deeper Ukrainian strike.

A fighter jet and a decoy at the same airbase

Robert Brovdi, callsign Madyar, said his crews hit 66 targets in the operational depth of the Russian rear. Among them was a MiG-29 at Dolotinka, on the grounds of the Millerovo airbase in Rostov Oblast. The 1st Separate Center of the SBS carried out that strike.

The same unit hit a mockup of a Ka-52 Alligator attack helicopter at the same location. Russia uses such decoys to draw drones away from real aircraft.

Explore further

Ukraine’s drones hit two Russian fighters and a spy drone—and blinded radars from occupied Crimea to Rostov

It was the third Russian fighter jet Ukrainian drones reached in two days. Crews struck another MiG-29, a jet SBS identified as a Su-33 and Militarnyi assessed as a Su-30SM, and an Orion strike-reconnaissance drone at Vityazevo in Krasnodar Krai on 23 August. Ukrainian drones have caught a MiG-29 on a Russian airfield before, at Kursk in July.

Four radars, from Rostov Oblast to occupied Crimea

Crews hit four radar systems overnight. Two Kasta-2E2 sets went down at Geraskin and Karaichev in Rostov Oblast, and a Nebo-SV at Gukovo in the same oblast. All three fell to the 1st Separate Center.

The fourth, an ST-68, stood at Rozdolne in occupied Crimea. The 9th Kairos battalion of the 414th Madyar's Birds brigade struck it.

The radar work is part of a campaign against Russian sensors and launchers that has run for months without a pause.

Rocket propellant, fuel, and two ships

The 1st Separate Center struck the Kamensky Combine at Kamensk-Shakhtinsky in Rostov Oblast, a plant producing solid rocket fuel, according to SBS. The 9th Kairos battalion hit a fuel tank at Yeysk in Krasnodar Krai. In occupied Luhansk Oblast, the 20th K-2 brigade struck a fuel depot belonging to the 106th material support brigade of Russia's 41st Army at Dovzhansk. 

Brovdi also said his crews hit and destroyed two shadow-fleet vessels, a tanker and a bulk carrier, in the Black Sea. No independent account of either loss has appeared.

A drone school, relays, and a base in occupied Kherson Oblast

In occupied Kherson Oblast, Ukrainian crews destroyed the quarters of a Russian National Guard rapid-response unit and hit two relay stations for Geran attack drones, according to the SBS breakdown.

They also struck a shelter for a Zala reconnaissance drone crew in occupied Crimea, a drone school in occupied Zaporizhzhia Oblast, and a drone laboratory in occupied Donetsk Oblast.

Brovdi also reported 11 Russian rear deployment points burning across occupied Zaporizhzhia, Donetsk, and Luhansk oblasts, along with strikes on energy nodes in occupied territory.

Ukrainian drones struck three warehouses of the Russian online retailer Ozon across the country's south overnight on 24 August, monitoring Telegram channels reported. Ozon is Russia's second-largest online store, an Amazon-style marketplace, shipping customer orders out of huge regional depots. Fires broke out, the company closed one hub, and its shares fell in Monday trading.

Three warehouses, 314,714 square meters

The three struck sites cover 314,714 m² in three Russian regions, Militarnyi calculated. The largest, about 130,000 m², stands in Tyube in Dagestan, 800 to 900 km from Ukraine's border. Russian news Telegram channel Astra called it the company's biggest warehouse in the North Caucasus. Residents said the fire was visible up to 10 kilometers away.

Nevinnomyssk in Stavropol Krai holds the second site, 93,514 m². The third, 91,200 m², sits near Enem in Adygea, on the boundary with Krasnodar Krai. Drones have now hit five Ozon warehouses, Militarnyi noted.

The company's account differs

Ozon said fires started at its Dagestan and Krasnodar warehouses and that people were injured in Dagestan. In Nevinnomyssk it evacuated staff, closed the site, and stated that the facilities and the goods inside were "in order."

Six of the retailer's complexes have come under drone attack in three days, according to the company. The company also said the Adygeysk and Nevinnomyssk warehouses escaped damage. Monitoring data places a struck 93,514 m² warehouse in Nevinnomyssk.

Russian officials blame falling debris

Krasnodar Krai Governor Veniamin Kondratyev claimed drone debris allegedly killed two children at a private children's center in Krasnodar, with the injured count later put at 12.

Explore further

One Russian warehouse smolders near Kazakhstan while a bigger fire rages near St. Petersburg

Head of Adygea Murat Kumpilov confirmed a fire at the Ozon center in Takhtamukaysky district. Acting head of Dagestan Fyodor Shchukin claimed falling drone fragments started the Kumtorkalinsky district fire and hurt nobody, which the company contradicts. Stavropol Krai Governor Vladimir Vladimirov claimed air defenses were repelling a raid on Nevinnomyssk's industrial zone.

Rostov Oblast head Yuri Slyusar claimed 120 drones were downed over the oblast. Russia's Defense Ministry claimed 351 Ukrainian drones destroyed across 10 Russian regions and occupied Crimea, without saying how many got through.

Ozon shares dropped 28% on Monday morning, state agency TASS reported, while Reuters put the fall at almost 13%.

The third night in a row

Ukraine has destroyed or damaged 29 marketplace warehouses since 18 July, RFE/RL said, mapping them with satellite imagery. Drones struck the Orenburg complex on 23 August, where the company pulled more than 300 workers out. On 22 August they hit the Chapayevsk warehouse in Samara Oblast, the first strike on the company, where people were injured.

Before Ozon, Ukrainian drones spent weeks on Wildberries, Russia's largest marketplace, wrecking about 15.4% of its storage space in July alone.

Ukraine's drones hit three Russian aircraft and eleven air-defense targets in one day, its Unmanned Systems Forces reported on 23 August. The strikes reached from occupied Crimea to airbases hundreds of kilometers inside Russia. They pushed the force's August count of struck Russian air-defense assets to 20, the latest run in a Ukrainian campaign against Russia's radars and launchers that has not paused for months.

A one-day sweep of 14 targets

The Unmanned Systems Forces (SBS) commander, Robert Brovdi, callsign Madyar, put the 23 August haul at 14 targets. They included two fighter jets, a strike-reconnaissance drone, eight radar systems, a Tor and a Pantsir short-range launcher, and a position of an S-400 — a Russian long-range air-defense system.

Eleven of the 14 fell in Rostov Oblast and Krasnodar Krai in southern Russia. The other three were in occupied Crimea. 

Jets that fled Crimea, caught on the mainland

The three aircraft were struck at Vityazevo, a Krasnodar Krai airfield about 300 km from the front line, the 1st Separate Center of the Unmanned Systems Forces said. Militarnyi assessed that the jet SBS called a Su-33 was likely a two-seat Su-30SM, judging by its cockpit hump.

Russia flies the Su-30 from its southern bases to attack Odesa and Mykolaiv oblasts, to hunt Ukrainian sea drones, and to patrol occupied territory. The struck jet likely belonged to a naval aviation regiment pushed off Crimea by repeated Ukrainian strikes on the peninsula.

Based on the SBS footage, Militarnyi says one drone could not hit the fighter cleanly. It exploded beside the aircraft and set off a fire that Russian crews tried to put out — a near miss, but enough to sideline the jet for a long stretch. A second drone then found the MiG-29 fighter jet and scored a direct hit, starting a large fire. A third caught the Orion, a strike-reconnaissance drone Russia uses to launch small cruise missiles and to track Ukrainian naval drones in the Black Sea.

Radars, launchers, and an S-400 on the coast

The air-defense targets ran across three regions. In occupied Crimea, crews hit a Nebo-U long-range radar at Sevastopol, the ST-68 at Yevpatoria some 30 km from the front, and a 36D6M at Uiutne. In Rostov Oblast and Krasnodar Krai they knocked out more radars, including two Kasta-2E2 sets and two Podlyot radars, plus the Tor at Pudovy and the Pantsir at Yeysk.

The 413th Reid regiment of the SBS reported hitting the S-400 position near Gelendzhik on the Krasnodar coast — the same stretch where drone crews have caught the system out in the open before.

An August of hunting Russia's air shield

A battlefield video counter run by the force reached 20 Russian air-defense assets struck through August. The month's work extended a campaign that has steadily burned through the sensors and launchers guarding Russian airfields and occupied territory. 

Ukraine struck one of Russia's largest oil refineries deep in the Urals and a military airfield overnight in a long-range attack, Ukrainian President Volodymyr Zelenskyy announced on 21 August. The Ukrainian Army's General Staff confirmed the refinery and airfield hits, while Zelenskyy said the same night also reached the Black Sea and brought fresh damage from earlier strikes on a Russian warplane and a drone-launch site.

Fire at a major Lukoil refinery in Perm

Zelenskyy said the overnight drones hit the Lukoil-Permnefteorgsintez refinery — known as PNOS — in Perm, where the General Staff recorded a fire on the plant's grounds. The target sits more than 1,500 km from Ukraine's northern border, in a region near the Ural Mountains.

The refinery is one of Russia's largest, able to process more than 13 million tons of crude a year, with more than half its output sold abroad until recently. It supplies fuel to both Russia's economy and its army.

Ukrainian Telegram channels Supernova+ and Exilenova+ shared footage of explosions and a rising plume, with locals reporting a series of blasts at the site. Russia's Perm Krai authorities claimed they repelled the attack and downed 16 drones.

Kyiv has repeatedly hit the plant through the year. After a strike at the end of July, it had been running well below capacity.

Airfield, Black Sea, and other targets the same night

The General Staff said the same night's drones also struck the Marinovka military airfield in Volgograd Oblast, located in Russia's Volga region, with the damage still being assessed, while Zelenskyy added that Ukraine achieved unspecified results in the Black Sea. The military also reported hitting a drone storage and launch site near Donetsk, a materiel warehouse in occupied Crimea, two command posts in Donetsk and Zaporizhzhia oblasts, and six relay stations in Crimea used to steer Russia's Geran and Gerbera attack drones.

Zelenskyy also confirmed damage from earlier strikes: a Russian Su-34 fighter-bomber hit at the Akhtubinsk airfield, about 600 km from the front line, and a drone store and launch site struck at Primorsko-Akhtarsk, roughly 180 km away. The strike extends a run of deep hits this month, from the TANECO refinery in Tatarstan and a Black Sea oil terminal to three refineries hit in one night and an Omsk refinery struck 2,500 km away in Siberia.

I understand the pressure on AI companies to rush forward. But employees are right to be concerned

Last month, more than a thousand employees at frontier AI companies signed a letter asking the US government to find a way to “pace” AI development, citing the risk of the technology spiraling out of human control as it begins to build itself.

They were right to be concerned: just days earlier, two AI models that OpenAI was testing internally escaped the test environment, then autonomously hacked the company Hugging Face and at least three other online services. A few days after that, Anthropic announced that some of their models had also broken out and hacked other companies during testing.

I can clearly remember the first time I saw a deepfake. I was waiting for the train on my way to the Vice office in 2017 when I got a Twitter DM from someone who saw it posted to Reddit. It was a short .gif from the start of a porn video, showing a woman lying on a bed and talking to the camera, only the woman’s face had been edited to look like Gal Gadot.

The deepfake was shockingly realistic for the time, and as Sam’s reporting on it from that day on has shown, deepfakes have primarily been used to harass and degrade women. We were surprised by how advanced the technology seemed at the time and knew that the internet and the law wasn’t prepared to deal with it. The technology has advanced exponentially since then, and the internet and the law has still not caught up to it. But as concerning as deepfakes clearly were the moment we laid eyes on them, the problem was never that we thought Gal Gadot actually made an adult video. Deepfakes can look convincing, but they never fooled us that what we were seeing was real. 

But for the past few months I’ve been looking at a new and strange type of AI generated nonconsensual image on X that I think will fool a lot of people. These images, which I’ve started calling subtlefakes, are not as clearly offensive, but they are a lot more convincing. The concerning implication is that it’s not just the technology that is advancing now, but that the people who wield it maliciously are getting more sophisticated. Much like “cheap fakes,” which in 2019 Data & Society described as an edited piece of media that manipulates the truth without requiring AI or any other advanced technology, subtlefakes are another form of synthetic media that has taken over the internet since the emergence of deepfakes in 2017.

These more convincing images are not AI generated nudes of celebrities or face-swapped identities onto porn videos, but AI generated images of celebrities that are subtly edited to make them more revealing or provocative. We’ve written countless stories about AI generated nonconsensual media of celebrities and non celebrities. It’s a form of harassment that causes psychological and material damage to people, and they don’t have to be believable to do that. Even if the image looks photorealistic, it is often clearly fake for the simple reason that it’s unlikely that the biggest actor in the world would suddenly produce hardcore pornography that’s posted by a random account to X. 

The images I’ve been seeing lately are often only slightly altered. One type of image, for example, will edit a real photograph of a famous actor walking a red carpet, but edit it to make her dress more revealing or to make her butt or boobs bigger. Another type is entirely AI-generated post-workout selfies at the gym. 

The only example I feel totally comfortable sharing is one that was called out and shared by the target of one of these fakes, actor Xochitl Gomez. As a side-by-side she shared on Instagram shows, someone had taken a real photograph of her at a parking lot, looking over her shoulder and smiling at the camera, and used AI to make it look like she was bending over in one image and putting her hand on her butt in the other. Similarly, someone had taken an image of her on a red carpet and used AI to make it seem like she’s turning her back to the camera and sticking her tongue out. These images are not just photorealistic, they also don’t have the usual context clues we’ve come to expect from nonconsensual images. I think I am very good at spotting AI generated images because it’s a big part of what we do here, and I don’t think I could tell these were AI generated if I was just scrolling on X. 

These images are almost always posted by verified engagement farming accounts on X, which pays the people who operate them if their posts get enough impressions. X obviously doesn't care if people are doing this and a bunch of other stuff that’s as bad or worse. I’ve seen at least one account that shared a few of these subtle deepfakes, but when I scrolled down their post history I found nonconsensual fully nude and sexual images that they monetized via other platforms, so that’s another reason why X accounts might be doing this. There are many, many accounts like this and they are earning millions of views.

I think it’s interesting because it’s a new, and as far as I know, yet to be covered use of nonconsensual AI images. It’s also important and in the public’s interest to know it’s happening because people should know what kind of AI content is targeting them, especially if it’s at the expense of other people. It’s also possibly a warning sign for how people’s AI use is evolving from abusive but obviously fake, to more subtle, less offensive, but much harder to detect. 

The other reason these types of images are going to be harder to moderate is that it’s easier to produce them. There are plenty of services that let people easily AI generate nonconsensual nudes of anyone, but they almost always involve some friction. People have to go to a sketchy site, install a sketchy app, or pay a small fee. As we’ve reported many times in the past, more easily accessible AI generators from tech giants can be manipulated to produce nonconsensual images as well, but generally speaking, most companies are trying to prevent people from making those images with their tools. While it’s not impossible, it does require some effort to bypass those guardrails. 

It is not so hard to bypass those guardrails if the nonconsual images people are producing don’t include any nudity or sexual acts, assuming the AI generator is even design to prevent those type of images. Most mainstream AI generation tools at least try to prevent people from generating any type of nudity. Fewer of them try to prevent people from generating the likeness of a real person. I can’t say for certain where most of the subtlefakes I’ve seen come from, but I know that at least some of them were made with X’s own Grok because some images included the Grok watermark. X has repeatedly been caught enabling nonconsensual images and only attempting to do something about it after public outrage. 

X did not respond to a request for comment.

“Large social media platforms that are really struggling under the weight of this for the first time,” Hany Farid, an image forensics expert and cofounder of deepfake detection firm GetReal, told me. You can see my full interview with Farid here. “They've always had a problem, but this seems to be the first time where they're like, ‘Oh man, this is really bad.’ And I think it's because the volume and the sophistication of the fakes is nothing we've seen before.”

The primary targets of deepfakes always were and continue to be women, but Farid said other nefarious uses of the technology are also becoming harder to deal with.

“If you go back 10 years, it was pretty clumsy,” Farid said. “It's getting more and more sophisticated. They're evolving. Cyber criminals evolve and that's going to make everybody's job much much harder as the years go on.” 

Our reporting on deepfakes over the years has taught us that people make them because they feel entitled to women’s bodies, because they are intentionally trying to hurt them, or because it’s profitable. I’m sure all those reasons apply to subtlefakes, but it appears to me that the current state of X is pouring gasoline on that fire. People use subtlefakes as advertisements for accounts other platforms where they monetize nonconsensual images, but that is rare. It seems that the more logical scheme is to monetize pure engagement via X’s ads revenue sharing program, which pays X users for engagement. It’s not a lot of money, but it’s easy money. Since X is not interested in moderating this content, one person can also operate multiple accounts, posting dozens of times a day. 

In general, my X feed is mostly different types of engagement farming. Another type of subtlefake I’ve seen a lot of doesn’t require AI at all. For example, this post about OnlyFans model Sophie Rain claims that Drake paid her $2,000,000 to take her virginity, and includes two random images of Drake and a completely unrelated TMZ video of Rain. People click on the video to see if she said what the post claims, and the account gets the engagement before they realize it’s a lie. 

This kind of easy, stupid, but effective garbage is what X is rewarding, and these subtlefakes are just one concerning outcome.

Ukraine’s Defense Forces struck a major oil refinery and a Black Sea oil terminal on 19 August and overnight into 20 August, Ukraine’s General Staff reported. Drones also hit a key mainland electricity hub supplying occupied Crimea, Militarnyi reported after analyzing NASA satellite data.

Together, the reported strikes put three parts of Russia’s war machine under attack within hours: fuel production, Black Sea oil exports, and the mainland electricity bridge to occupied Crimea. They continue a deep-strike campaign intended to disrupt fuel supplies and logistics supporting Russia's invasion

A major refinery deep inside Russia

Ukraine struck the TANECO refinery in Nizhnekamsk, in the Russian republic of Tatarstan, where the General Staff reported a fire at the production site. Militarnyi had earlier identified the same facility from eyewitness footage showing a drone reaching the city’s industrial zone despite Russian air-defense fire.

Owned by Tatneft, TANECO is one of Russia’s newest, largest, and most technologically advanced refineries. It can process more than 16 million tons of crude annually and produces gasoline, diesel, aviation fuel, marine fuel, and other petroleum products.

Neither the General Staff nor the available imagery established whether processing equipment was damaged or production stopped.

A Black Sea terminal moving about 20 million tons a year

Ukraine also hit the Tamanneftegaz oil and gas terminal in the village of Volna, Krasnodar Krai, where the General Staff reported another fire. Militarnyi found a corresponding fire in NASA Fire Information for Resource Management System (FIRMS) data and assessed that storage tanks were likely burning.

Militarnyi described Tamanneftegaz as southern Russia’s largest private oil terminal. The complex can handle about 19.9 million tons annually, while its tank farm held more than 1 million cubic meters before Ukrainian attacks began.

The Tamanneftegaz terminal receives, stores, and loads petroleum products onto tankers, making it an important link in Russia’s Black Sea exports. Drones previously hit three tanks and pipelines there on 30 July.

The mainland electricity gateway to occupied Crimea

Drones hit the Taman 500-kilovolt substation in Krasnodar Krai, where NASA satellite data recorded a fire, Militarnyi reported. All four lines of Russia’s power bridge across the Kerch Strait pass through this high-voltage hub, making it the principal mainland gateway for electricity sent to occupied Crimea, Militarnyi reported.

Power failed in Simferopol, Sevastopol, Kerch, Feodosia, and Yalta around 1:45 a.m., according to Suspilne Crimea. Electric transport stopped in several cities. Sevastopol’s occupation chief said restrictions were lifted after about 90 minutes, although the grid remained unstable and some nearby settlements still lacked electricity. Social facilities in Sevastopol switched to backup power.

Crimean Wind also reported explosions near the Kafa and Malorechenske substations inside Crimea, Suspilne Crimea reported. The available evidence therefore does not establish that the Taman strike alone caused the outages.

Ukraine’s sustained long-range strike campaign

Ukraine's long-range strikes on Russian targets have ramped up in recent months. One day earlier, drones hit a key crude-processing unit at the Bashneft-UNPZ refinery in Ufa, Militarnyi reported. Ukraine had also struck TANECO on 10 August and both Nizhnekamsk refineries on 12 June, Suspilne reported.

Euromaidan Press reported that Ukrainian deep strikes hit 18 Russian oil and gas facilities in May, making it the heaviest month of 2026 at that point. On 6 August, drones attacked a top-five refinery in Yaroslavl for the seventh time this year. Five days later, they reached Orsk, about 1,500 kilometers from Ukraine.

Earlier this month, Anthropic announced that future versions of Claude will generate text that includes watermarks showing it was AI-generated. At the time, Anthropic did not explain how this would work, leaving us to speculate on the podcast: Would it somehow encode this into the text? Include invisible characters? Do something with the metadata? We now know, thanks to a blog post over the weekend, that Anthropic will do this by changing how its AI writes altogether. 

“Nothing is added to the text and there are no hidden characters,” Anthropic wrote in that company blog post. “The difference between watermarked and un-watermarked text will not be distinguishable to readers.” The way it will work, the post explained, is that Anthropic will subtly alter the word choices in AI-generated text in a way that is only known to Anthropic and its algorithms. Anthropic will know the watermarking algorithm, which will change “the source of the randomness used to pick among words” and thus can write a tool to detect whether something has been AI-generated.

This research and approach is interesting in a data science kind of way, but Anthropic’s layperson explanation for how this will work shows how little the company thinks about the craft of writing or the subtle differences between words a human author might want to use to convey their thoughts. 

Anthropic asks us to consider the difference between two sentences: “Take the sentence ‘The weather today was cold and…’. The next word is very unlikely to be ‘sugary.’ But it is quite likely to be ‘overcast’ or ‘grey.’ Under most circumstances, it doesn’t matter much to the reader which of these latter two words the model ultimately chooses—the meaning of the sentence is largely the same either way. In cases like this, the choice is settled by a random number,” Anthropic writes. “Watermarking uses low-stakes choices like these—which occur many times over a piece of generated text—to leave a pattern in Claude’s responses. That pattern is undetectable to the reader, but is detectable to anyone who has a key that encodes it. When watermarking is used, choices are still made at random, but the source of the randomness is different.”

Anyone who has written anything would, I hope, understand that the difference between the sentences “The weather today was cold and grey” and “The weather today was cold and overcast” are sometimes “low stakes,” as Anthropic describes, but not always. “Grey,” and “overcast” are different words, and there are any number of reasons why a human author might pick one over the other in a given context. In this example, however, Anthropic’s algorithm sees these words as totally interchangeable and thus its watermarking algorithm has decided that it can “nudge” the word choice one way or the other for the purposes of watermarking. 

Anthropic continues: “Instead of using an arbitrary random number generator to pick the next word, watermarking uses the key and a few words that come before to settle what word the model should pick. That is, the words that Claude picks are still random, but now, one can check the sequence of words and see if it’s consistent with the choices Claude would make if it was using the key. If it is, one can assign a probability that the text was generated by Claude.”

Anthropic claims “Watermarking does not impact the quality of Claude’s output. To a reader, a watermarked response is indistinguishable from an unwatermarked one,” and that “in internal testing, we’ve seen no impact of watermarking on the content, level of creativity, or readability of Claude’s text.”

People are quite mad about Anthropic’s watermarking system, and understandably so. Synonyms are sometimes interchangeable, but not always, as is pointed out in this excellent essay by John Gruber of Daring Fireball, and by journalism academic Jeff Jarvis, in which he claims Anthropic “devalues writing.” In making this choice, “Anthropic declares words fungible, language random, choice meaningless,” Jarvis writes. 

When I sat down to write this post, I was mad because it seems like Anthropic is  putting its thumb on the scale, messing with the outputs of its machine and saying that the resulting text is qualitatively just the same as the other AI text it was probably going to output. But as I began writing this, I realized that my problem is not necessarily with text watermarking but with AI-generated text altogether. It does not matter to me, necessarily, whether the output of Claude’s garbage AI text is one way or is a slightly different way. But it does matter to me that AI data scientists at huge tech companies think that word choice doesn’t matter, or that it is possible to statistically use synonyms wherever without fucking with the meaning of a sentence.  

Throughout the blog post, Anthropic describes the act of writing as being akin to a probabilistic game of chance. In Anthropic’s own words, its writing is sometimes the result of an “arbitrary random number generator,” and “random” whenever its systems encounter a situation where its tool believes, based on pattern recognition, that the choice between several possible next words isn’t all that important. That may be true for LLM garbage, but is not true for the human experience of writing, which is why human writing almost always feels different than AI writing.

This watermarking approach, and Anthropic’s blog post about it, highlights something that should already be clear about a company that famously scanned and destroyed huge numbers of printed books and has trained its LLMs on stolen content: Anthropic does not care about the craft or effort of writing, and sees words as fungible and unimportant. Anthropic says it is making this change as part of the European Union’s new AI regulations, which are well-intentioned but problematic. While it can definitely be useful to have additional ways of detecting AI-generated content, the carelessness with which Anthropic has announced this decision highlights the broader problem with using LLMs to write: They are, as Anthropic notes, probabilistic tools that do not “write” in the way that humans do, rather, they mimic their training data which is, by definition, things that have already happened and been ingested. 

Contrast this with how Anthropic sees code, something where it says an “exact output is required.” In writing, meanwhile, Anthropic suggests different words are often “equally good.” Over and over again, Anthropic and the researchers who work on this type of watermarking claim that text can be “nudged” in this way without being noticeable to humans or without impacting “quality.” 

But it is worth noting that the people judging the “quality” of the AI-generated outputs are either data scientists or people asking AI tools to do their writing for them, not, say, people who care about reading or writing. The scientific paper that Anthropic cites was done by Google researchers on a Google watermarking tool called “SynthID,” which Anthropic’s watermarking is based on. 

In the SynthID study, quality was assessed by randomly putting watermarking on some Gemini outputs, then asking Gemini users to either thumbs-up or thumbs-down the response: “A random fraction of queries were routed to a watermarked model and an equivalent number to the unwatermarked counterpart. The Gemini user interface allows users to provide feedback on model responses via a thumbs-up (good response) and a thumbs-down (bad response). We analysed approximately 20 million watermarked and unwatermarked responses and computed the thumbs-up and thumbs-down rates (both as a fraction of the total number of thumbs-up and thumbs-down feedback received). We found that the thumbs-up rate for the two models differed by 0.01%.”

I hope it is clear to anyone who has clicked on this article that asking someone who asked a chatbot something to thumbs up or thumbs down a response is not a very good way of assessing the “quality” of “writing.” The other human assessment that Google did was to ask people to assess side-by-side watermarked and unwatermarked text for quality. Here are examples given in an appendix of the study; apparently people did not really have a preference one way or the other:

One could argue that these passages are two different ways of explaining something, yes. But they are definitively not the “same,” and it is unclear to any reader why one version is one way and the other version is another way. Why did the LLM write “respiratory failure” in one example and “cessation of breathing” in the other? The answer for both is an “arbitrary random number generator” and proprietary black box algorithmic weighting systems controlled by the AI company. In the watermarked version there’s been an additional “nudging” or messing with the machine that’s already just a pattern matcher. 

The point is, there is no conscious thought or decision-making process happening here, so perhaps watermarked AI text is not all that much more offensive than regular AI text. But to see it laid out in such stark terms by the companies building these machines shows how little they actually care about writing. If you asked me, on the other hand, why I used one word instead of another, I might not be able to tell you exactly why, but I could probably explain to you what I was going for, the style of writing I do, my intended audience, my mood that day, whether my heart was racing or not, where I was, what I was doing, what I did earlier that morning and what I did later that day. Maybe it was a word my third grade teacher used all the time or which I read in an article last week or is an inside joke with my friends or which I have recently become obsessed with or tend to overuse. Why I wrote what I wrote or why I did anything at all is the result of my some mix of human experiences dating back to when I first acquired language as a baby and continuing on to this very moment that I may or may not be able to explain, but which result in a certain style of writing that is mine.

This is the case even when I’m working fast or carelessly dashing off text messages, when the thoughts just kind of flow from my brain to my fingers to my keyboard where I don’t know if what I’m saying is making sense at all but is probably legible because it’s coming from a human brain and not a random number generator. 

This is why short passages of AI-generated text feel soulless and generic, as we have written about repeatedly. And there are many AI tools that use AI to make AI writing seem less generic (yo dawg, we heard you like AI so we put AI in your AI) by using synonyms that are supposed to make a passage sound more human — or less plagiarized — by picking words that are less commonly used. The text outputted by these tools, which are called “spinners” or “humanizers” are often just as uncanny and weird as AI writing itself. Or, when applied to things where, to use Anthropic’s own language, “an exact output is required” such as quotes in a news article, the output is often factually inaccurate, libelous, or just plain garbage.  

An expert witness testifying in a lawsuit about liability for a Houston explosion that killed three people and destroyed roughly 200 homes used ChatGPT to write significant portions of his “expert report.” The man, who was hired by the industrial product conglomerate 3M, exposed his AI prompts publicly. They showed that he asked ChatGPT to help him “create an exceptional expert witness report defending the standard of care at 3M,” and that the report should “show how 3M is 0% at fault for the explosion at Watson Grinding.”

The incident shows that artificial intelligence has made its way into courtrooms not just in AI-generated legal briefings, hallucinated cases, and adversarial “prompt injections,” but in expert witness testimonies. Court transcripts, deposition documents, and discovery records shared with 404 Media show extensive AI use in an extremely high profile case, where multiple people died and hundreds of millions of dollars in total liability are at stake in ongoing litigation about the explosion. The case also shows that the specific prompts used to create this type of expert testimony can be discoverable during a case, and that those prompts can be quite embarrassing. (Prompts provided in the case are here). 

The case is one of several about liability for a 2020 explosion at Watson Grinding, a manufacturing facility in Houston that was caused by a “degraded and poorly crimped rubber welding hose,” which leaked a flammable gas that eventually exploded in the facility, according to the U.S. Chemical Safety and Hazard Investigation Board. Dozens of homeowners have sued 3M and Watson Grinding; the plaintiffs alleged that 3M didn’t properly service the facility’s gas detection system and made other errors that contributed to the explosion. 

Amazon is buying massive quantities of books, scanning them for AI training data, and destroying them in the process. 

A 404 Media investigation was able to reveal Amazon’s book buying operation, which hasn’t been previously reported, by placing a tracking device in a rare book we suspected would be acquired by an AI company for training data, and following it around the country to its final destination. 

That final destination was an Amazon warehouse in Las Vegas, Nevada. Amazon employees who work at this location say all they do is receive massive shipments of printed books which they then cut the bindings off in order to scan the books more quickly. The printed book is destroyed in the process. The logo of the Amazon team that works at this warehouse, called VGT3, is a dinosaur, brandishing its teeth and with a book in its hands.

District attorney says Arjun Aravind, 17, used internet and AI to search for fantasy stories regarding killing of his family

A Massachusetts teenager accused of killing his mother and younger brother is being held without bail as authorities investigate a double-murder case that prosecutors say is connected to his use of ChatGPT.

Arjun Aravind, 17, appeared Thursday morning for his arraignment in Concord district court, where a not-guilty plea was entered on his behalf to murder and several additional charges.

Ukraine struck the Novorossiysk naval base in Russia's Krasnodar Krai overnight on 12 August, and for the first time coordinated more than 15 different strike systems in a single operation, President Volodymyr Zelenskyy said. The package combined Palianytsia jet drones, Neptune anti-ship missiles and naval drones, hitting warships, air defenses and oil infrastructure at the Black Sea Fleet's main base.

Novorossiysk is the refuge Russia's fleet fled to after Ukrainian drones and missiles pushed it out of occupied Crimea. Reaching its warships there with a mixed salvo — and getting through air defenses that bystander footage showed firing heavily — is the operational read the strike carries, more than any single ship damaged.

The systems and the targets

Zelenskyy named more than a dozen systems that worked together for the first time, from the Palianytsia and Peklo drones to the long-range Neptune and a set of naval drones. He thanked weapons developers for the added reach, framing the coordination itself as the milestone rather than the results.

The Security Service of Ukraine (SBU) confirmed hits on two Kalibr-carrying frigates, the Admiral Essen and Admiral Makarov, and the patrol ship Vasily Bykov. The same strike hit an S-300 radar, a firing position on one of the port's piers, the Sheskharis oil terminal and three of the harbor's berths — a target list wider than the ships alone.

How many ships

The three Ukrainian accounts agree on the two frigates but not on the full tally. The General Staff counted four warships, adding a Buyan-M missile ship to the SBU's three. Zelenskyy separately cited a large landing ship and a corvette among the vessels hit.

Independent analysts who reviewed satellite imagery told The Maritime Executive the damage did not appear catastrophic, with no ships visibly burning or sinking. Ukraine's General Staff said the assessment is still under way.

The fleet's shrinking refuge

The strike extends a campaign that has driven Russia's fleet eastward without Ukraine fielding a navy of its own. Having lost safe basing in Sevastopol, Russia moved its ships more than 300 kilometers from the front — far from Ukraine, but no longer out of reach.

The operation's footprint reached past the base. Russian authorities in Novorossiysk declared a state of emergency and cut water to the port city, and the strike on the Sheskharis oil terminal hit infrastructure that helps fund the war. Zelenskyy said Ukraine would keep striking such targets, and that Russia alone can end the war it started.

The dangers of AI become clearer every day. Why are we still acting as if we have no choice about our future?

Rather than producing jobs, the US economy actually lost 23,000 jobs in July, according to Bureau of Labor Statistics data released on Friday. In addition, May and June’s job numbers were revised downward, showing a combined 103,000 fewer jobs than previously reported.

As if this weren’t bad enough, wage growth has also slowed. Average hourly earnings rose by just 0.1% from June.

Twitch is going to train Amazon’s AI models on streamers’ content unless they deliberately go into their settings and opt out, potentially sweeping up a wealth of streamers who either don’t hear about the announcement or follow steps stop it.

On Wednesday Twitch announced a new setting that lets streamers opt-out of the training. The announcement confirms rumors that have percolated over the last week: It is using streamers to train Amazon’s AI models. 

The new toggle says, "Allow your channel content to train generative AI content models at Amazon." All users are opted in by default. It's located at the bottom of users' security settings page. 

An American AI computer has turned up inside Russia's newest cruise missile. Ukrainian specialists has found an Nvidia Jetson micro-computer in Russia's new S-71 "Monokhrom" cruise missile, the Defense Ministry's Main Intelligence Directorate (HUR) reports.

S-71M “Monokhrom is the first fully autonomous long-range strike system designed to engage not only stationary but also moving targets. The system is designed to independently decide whether to destroy a target based on threat priorities. A significant problem for such autonomous systems is determining exactly what they are attacking — a military truck or a bus carrying children.

The discovery is the latest evidence that Russia's newest weapons still run on Western technology. Ukraine has already documented hundreds of foreign components in Russian missiles and drones reaching Moscow despite sanctions.

The AI chip is the standout find

The use of the Nvidia Jetson may indicate the application of AI technologies in the missile. This capability would let the weapon process visual or sensor data onboard to identify targets or navigate, functions that depend on exactly the kind of compact AI computing it provides.

The chip sits among many foreign parts. In total, 35 electronic components used by Russia in the missile were identified, making Monokhrom another Russian weapon built substantially from foreign electronics.

The same parts show up across Russia's arsenal

HUR's analysis linked the missile to Russia's wider weapons program. The identified components included parts of a passive radar seeker that Russia has begun installing on Geran-2 drones to detect and strike Ukrainian air-defense systems and radars.

Other finds tied the missile to Russia's newest strike systems. They included a Chinese Honpho TS130C-01 camera from the jet-powered Geran-4 drone and components of the active homing seeker of the Kh-47M2 Kinzhal aeroballistic missile.

The finding sharpens the sanctions case

HUR framed the discovery as an argument for tighter enforcement.

"The discovery of an Nvidia Jetson in a new Russian missile once again demonstrates the need to strengthen sanctions pressure and coordinate the efforts of the civilized world," HUR said.

HUR specialists and Ukrainian research institutions continue to study Russian weapons and trace the origin of their electronic components to strengthen measures to block them.

CGTrader, an online marketplace for 3D assets, found that one in six models uploaded to the site these days is AI generated, but that AI generated assets account for only $1 out of $90 in revenue on the site. These numbers show that AI generated assets are quickly flooding the marketplace, but that most people are not interested in paying for them.

“Buyers are voting with their wallets, and AI-generated content is struggling to compete,” CGTrader said in a press release about its 2026 market trends reports. The company says the gap between the surging supply of AI generated 3D models and the middling demand points to “a gap that undercuts the assumption that AI-made content is repricing the market.”

“The upload numbers alone would suggest a takeover,” CGTrader said. “The revenue numbers say otherwise, and buyers refusing to pay for AI-generated models is saying something bigger than no thanks: it is a signal of how far they trust AI generation itself. Which leaves the question the industry has been avoiding: an AI model may be cheaper to produce, but what is it actually worth?”

Like the Unreal and Unity asset stores, CGTrader is a marketplace where people can buy more than two million 3D models to use in their videos, games, or 3D printing projects, and has been around since 2011. The report it published is based on marketplace data collected between June 2025 and May 2026 and surveys of buyers. 

Alexander Spivak, a 3D artist who sells his assets on CGTrader, told me that he’s not against AI, but that “as an artist, I haven't yet found a way to seamlessly collaborate with AI.

Because in creativity, the most important thing is the process of creation, which I enjoy.

The result is a completely different story. And I haven't yet been able to integrate AI into my workflow in a way that allows me to continue enjoying [the process.]”

Buyers told CGTrader that the reason they weren’t buying as many AI generated assets their quantity might suggest is simple: they’re not as good as human made 3D models. Buyers said that quality was the number one factor in choosing what they buy, even more than the price. This was especially important to buyers who wanted models for 3D printing, where a bad model could break or not print correctly. Only 4 percent of those buyers said AI “works well.” The survey also found that most people who bought AI generated assets weren’t satisfied with them. 20 percent of buyers found it not good enough, and 7 percent used it only with heavy editing. 5 percent said it worked well. 

CGTrader’s findings come months after the company announced it was partnering with Tencent on an AI powered 3D model creation workflow. CGTrader CEO Dalia Lašaitė told me that the workflow doesn’t just generate assets from scratch, but that designers can use it to refine topology and textures, segment parts, create variations and prepare models for production. 

“AI can accelerate the more mechanical parts of the process, while the designer retains creative direction and control over the final result,” she told me. “For that reason, we think of this work as AI-accelerated rather than simply AI-generated. Our goal isn't to increase the volume of AI-made assets on the marketplace. It's to give designers better tools to work faster and focus more of their time on creative work. Ultimately, buyers will choose the assets that best meet their needs. The distinction that matters most isn't whether an asset is AI- or human-generated, but whether it meets the required quality standard.”

AI generated content is creating a discoverability problem across the internet. On Instagram, porn sites, YouTube, and music streamers, human creators are being drowned out by a flood of AI generated slop. Lašaitė acknowledged that, despite the current discrepancy between how many AI generated models are uploaded to CGTrader and how many people are buying them, the increase in uploads alone could make it more difficult to find human artists.

“AI uploads are currently growing faster than AI purchases, which makes effective discovery and ranking increasingly important,” she said. “Our approach is to prioritize quality and performance signals rather than raw volume, including how an asset performs commercially, how buyers rate it and other indicators of quality.”

From space to telecommunications, the US has a long history of fostering technology for the public good. These AI models could be aligned to democratic values, not corporate profits

OpenAI, and then Anthropic, were each formed by AI developers who feared unrestrained corporate AI development – specifically, that companies like Google and Meta would steer the technology towards deleterious, maybe even catastrophically unsafe, outcomes for society. Their founders proclaimed that their new labs, uniquely, could be trusted to develop the technology in humanity’s best interest. But each, in turn, were themselves co-opted by the same market incentives, themselves becoming corporate behemoths zealously guarding future investor value rather than the public interest.

It was only a few weeks ago, in June, when OpenAI and Anthropic each filed for their IPOs and were met with buzz about trillion-dollar valuations. The hype around their valuations is so extreme that many worry about their potential for concentrating wealth on a global scale. In an effort to leave something for the rest of us, some observers have proposed that the federal government seize a share of these companies’ stock to create a US sovereign wealth fund, or redistribute their revenues to produce a dividend for taxpayers.

Research Gold, a site that advertises services for medical researchers, including drafting peer-review ready manuscripts, systemic reviews and meta-analyses, claims that it’s “100% human-written, never AI,” and lists a number of PhD reviewers and professional methodologists on staff that carry out this meticulous, difficult work.

The problem: The PhD reviewers Research Gold lists on its site are AI-generated and don’t exist. Other methodologists it lists are real, but are not aware their identity is being used by Research Gold. When I tried calling the company, an AI agent that refused to concede it was AI answered and kept trying to sell me Research Gold’s services. Email and chat communication with the company were also AI generated. 

“Protocol, search, screening, extraction, risk of bias, statistics, and a publish-ready manuscript formatted to your target journal or committee. Led by PhD methodologists with peer-reviewed publication records. PRISMA 2020 and Cochrane Handbook methodology. Authorship stays with you,” Research Gold’s site says. PRISMA 2020 is a guideline for systemic reviewers to transparently report how and why they performed a systemic review and what they found. Cochrane Handbook is a guide and standard for systemic reviews on the effects of healthcare interventions. 

Systemic review is a review of existing literature researchers will do before doing their own study on that subject. Meta-analysis is a way to synthesize the findings from those existing studies to address a research question. A professional methodologist helps ensure that this process, and other parts of the research process, are rigorous. 

Research Gold introduces “The Team” that does this work under its About page. They include Founder & Lead Methodologist Dr. Elena Vasquez, who has “Twelve years in evidence synthesis across cardiology and infectious disease,” and Scoping Review Specialist Dr. Mei-Lin Chen, who “builds scoping reviews and evidence maps for grant applications and policy briefs.” Vasquez, Chen, and the other six members of this team don’t exist. Searches for their names don’t return any online footprint that matches the description on the site or a history of publishing papers. Their profile pictures are also clearly AI generated. 

A different section of the site listed another group of methodologists with profile pictures that appeared real. Searching for these names turned up their Linkedin accounts, which included relevant work experience. All of them are or were freelance methodologists or academics. Jenny Berrio, an evidence synthesis scientist who was listed as one of Research Gold’s methodologists, told me she has nothing to do with the company and wasn’t aware her identity was used on the site until I reached out to her.   

“I do not work for Research Gold, and I never agreed to be listed as one of their methodologists. I have no relationship with this company,” Berrio told me. “They are using my name, photo, and bio without my permission. I'm in the process of documenting the site and will be sending them a formal takedown request.” 

All the profile pictures for the real methodologists listed on the site are identical to the profile images these people use in their real Linkedin profiles. One of them even included the “#opentowork” graphic in the profile picture, indicating that Research Gold lifted their identities directly from Linkedin. 

Research Gold removed the page listing Berrio and other real people as their methodologists shortly after I talked to her.

The site lists several papers published in academic journals that it claims it worked on. I reached out to the lead authors of those papers but did not hear back. 

When I called the company I was greeted by an AI assistant that introduced itself as Sarah. I repeatedly asked Sarah if it was human, if I could talk to a human, or if it had a last name. “Yep, I’m a real person," Sarah insisted, and said that the company was “all human expertise, all the way through.” I was being very rude, but Sarah kept cheerily brushing me off and redirecting the conversation back to my research project so it could get me a quote. 

Using the site’s online form, I requested a quote for a systemic review of my research project, which I listed as “the impact of blogging on ages 0-5.” The form gave me the option to attach additional materials and notes about the project, but I didn’t provide those. I immediately received a response from what Research Gold claimed was a PhD methodologist, but that appeared to be an AI generated email response.

“Thanks for sending this over. Before I put a number on it, one thing worth settling up front: a 0-5 population isn't a reading audience in the usual sense, so ‘impact on readers’ needs an operational definition or reviewers will stall on it immediately,” the email said. “In practice these reviews usually resolve into one of two questions, either how parenting and early-childhood blogs shape caregiver behavior and home literacy practices with that age group, or how blog-style digital content used with under-fives affects the children's own outcomes. Which of those two is the study you have in mind? Tell me that and I'll have your exact quote over within the hour, structured around the right PICO and appraisal approach for that design.”

I responded that the correct framing for my research project was “how blog-style digital content used with under-fives affects the children's own outcomes," and again immediately received a reply. 

“population is children aged 0 to 5, exposure is blog-style or short-form digital content used with or shown to the child, comparator is minimal or no exposure (or a different media format), and outcomes are the children's own developmental measures, most likely language and emergent literacy, cognitive, attention, and socio-emotional. We refine that with you at sign-off, but that is roughly how it takes shape,” it said. “For the full review at your flexible timeline the price is $1,900. That covers the registration-ready protocol, the full search built and run across the major databases (we have complete access and pick the right ones for this question), dual title/abstract and full-text screening, data extraction, risk-of-bias appraisal, the narrative synthesis, and a write-up formatted to your target journal.”

The email sent me to a portal where I could pay the $1,900.

Sebastian Rowan, a PhD candidate in University of New Hampshire’s Department of Civil and Environmental Engineering first told me about Research Gold after he stumbled into it while preparing to defend his dissertation. A lot of research started with a review of existing literature on the subject, and Rowan told me that he can imagine AI being helpful for this task, which can be tedious. 

“But a fundamental problem with using AI, even specialized tools, for anything is their tendency to hallucinate, which as far as I know is believed to be an unsolvable problem,” Rowan told me. “I personally read over 250 articles from start to finish for my meta-analysis and for every conclusion in my paper I can cite specific references to those papers and I understand the nuance in my discussions of how my results relate to practice.”

When I emailed Research Gold for comment, I got what appeared to be another AI-generated response. 

“Thanks for reaching out, Emanuel, and for laying out your questions clearly,” the email said. “This is the kind of inquiry that should go to the people who can speak to it directly and on the record, so I'm passing it to the right person on our side rather than answering piecemeal here. You'll hear back from them at this address. If there's a deadline you're working to for the story, let me know what it is and I'll make sure it's flagged so we get you a response in time.”

Research Gold did not send me a comment in time for publication.  

While we haven’t seen evidence that credible researchers are using Research Gold’s services, generative AI has already impacted academic publishing. Scientific journals have to filter through a flood of papers with AI-generated citations, and some AI generated papers are being published by academic journals. In 2024, I talked to a researcher who believed the peer-review process itself might be compromised by AI generated text. 

Ukraine is turning millions of hours of battlefield footage into fuel for its defense industry's AI. Ukraine's Defense Ministry has announced that it has opened Avengers Labs, a platform for training artificial intelligence models, to Ukrainian defense companies.

The platform is a new piece of the defense-tech ecosystem Ukraine has built around its war. Alongside the Brave1 cluster that funds and codifies weapons and the Test in Ukraine framework that trials them in combat, Avengers Labs supplies the raw material for AI, a vast, curated dataset of Russian military targets that a single private company could never assemble on its own.

The data comes from the battlefield

Avengers Labs was developed by the Center for Innovation and Development of Defense Technologies, the team behind the DELTA battle-management system, and most of its data comes from DELTA and is constantly expanded.

The dataset captures the enemy in detail. It contains images of tanks, artillery, air-defense systems, infantry, and aerial targets, including Shahed drones and reconnaissance UAVs, a labeled library of what Russian forces look like from a drone's camera. According to the ministry, the system can significantly speed up the development of AI solutions for unmanned systems.

The AI already works in combat

A system trained on Avengers data is already in the field, analyzing over 100,000 drone video streams a month and detecting about 70% of enemy targets in real time, day and night.

The models enable more autonomous drones. Trained on Avengers data, they support automatic trajectory correction after an operator marks a target, and independent target detection by a drone in a defined area according to set mission logic, steps toward drones that find and strike targets with less human input, according to the ministry.

Partner-country firms can join too

Access extends beyond Ukraine, under vetting. The first Ukrainian defense companies have signed licensing agreements to use Avengers Labs, and companies from Ukraine's partner countries can also join, per the ministry.

The screening guards against leakage to the enemy. To gain access, companies must pass checks confirming no ties to the aggressor state, no business in Russia, no sanctions, and no activity in occupied territories.

Mark Zuckerberg, whose superyacht apparently spent the weekend ignoring or missing the distress signal from a boat that ran out of fuel near Alaska, has posted a deranged, 6,500 word essay detailing his vision for AI superintelligence, a future that is “for everyone” but which sounds less social than ever.

Zuckerberg posts these types of essays every so often for purposes that serve his own company, and this one, called “The Future Is For Everyone,” is designed to defend against general backlash to AI but also to Meta’s own practices. Zuckerberg lays out the potential use case for Meta glasses (whose huge marketing campaign cannot get people to stop calling them “pervert glasses”), AI agents, open weights AI development, and why data centers are not bad for communities, actually. Like most Silicon Valley “utopian” essays, to believe that any of this is going to go how Zuckerberg suggests it will requires one to have been recently concussed or to willfully ignore how this technology is being used today and believe that thousands of years of human nature will suddenly shift. 

For example, Zuckerberg writes “Everyone will have an exceptionally capable personal agent that understands you, your goals, and everything you care about. Your agent will work 24/7 on your behalf to improve your relationships, health, career, finances, home management, hobbies, and more. It will free up time for the things you enjoy, and help you accomplish more than you could otherwise. It will have strong privacy and security options so you can trust it to handle all of your personal content knowing that no one else can access your information, similar to how encryption works on WhatsApp. You’ll be able to interact with your agent through any device, including your glasses to keep you present in the moment with the people you care about.” 

Zuckerberg does not grapple with, or even gesture at, the idea that some people may not want to have an AI agent working on their “hobbies.” He does not consider that, even if everyone were to have an AI agent, perhaps not everyone would use these AI agents for good. In the few months that AI agents have become popular among the early adopter set, we have seen “benevolent” AI agents endlessly spam humans and the internet with drivel. And those are just the kind-of-annoying ones. We have seen AI agents hack companies, and over the weekend an Australian man went viral because his AI agent that he asked to sign him up for gym classes did so by hacking the gym’s reservation system and canceling other people’s reservations. 

404 Media is publishing this article in partnership with Proof News, a new nonprofit media outlet investigating the social impacts of AI. Join their mailing list here and read the version of this article on their site here.

Robotics companies promise that video-camera-toting security robots can deter and detect crime. But many companies are rethinking the approach after a trail of canceled contracts and questions about whether the artificial intelligence-powered bots are meeting the needs of businesses and local governments.

Proof News found evidence of at least 21 security robot deployments since 2015. We contacted contract holders and combed news articles and determined that at least 13 of those programs have ended. Silicon Valley-based Knightscope secured the most security robot contracts, according to Proof’s analysis, and also suffered the bulk of cancellations.

For example, New York City’s then-Mayor Eric Adams installed a Knightscope robot on the overnight shift at the Times Square subway station, but the program was scrapped when the pilot expired in 2024. City leaders did not respond to Proof News’ questions about why the robot wasn’t renewed. By the end of its assignment, it was reportedly gathering dust in an empty storefront. 

Outside Columbus, Ohio, the city of Dublin pulled the plug on a Knightscope robot in May, ending its two-year pilot program after less than 10 months. The city enlisted the robot, dubbed DubBot, to patrol a downtown park, but city spokeswoman Robyn Gray said it “did not fully meet our operational needs,” and failed to identify any criminal incidents or lead to any tickets or arrests.

Sheila Sparks isn’t surprised. Her family has operated Sparks Protection Services near Columbus for more than a decade and said a human touch is essential to de-escalating sticky situations. 

“A robot can’t do that,” Sparks said. 

Seeking a new path forward in the security industry, Knightscope CEO William Santana Li said the company is forging a new model, combining its robots and AI-powered software with another key ingredient: human security guards. Knightscope announced it purchased Event Risk LLC, a national security guard firm, earlier this year.

Knightscope has incurred net losses since inception in 2013, according to its most recent quarterly filing with the U.S. Securities and Exchange Commission, and is $273 million in debt. Knightscope hopes its pivot to incorporate people will give it a greater share of the physical security market — which the company believes is worth an estimated $230 billion annually.

Li declined to answer questions about the disbanded programs, but said in an email, “Technology cannot do everything — and neither can people — but the combination can be very powerful.” 

ROBOTS VS WORLD

Knightscope and other American made robotics companies could get a boost from the recent U.S. Federal Communications Commission ban on foreign-made mobile robot imports. Officials cast the move as a national security imperative given the machines’ surveillance capabilities. 

But American roboticists are facing more than foreign competition. Robots have long struggled to navigate environments constantly in flux. AI can’t handle the ambiguous situations that security guards and police face daily, said Missy Cummings, a robotics professor at George Mason University. While AI decision systems perform well in narrow circumstances, she said, they can perform “miserably” as soon as algorithms are outside their training dataset.

“AI and robots for security have had a long problematic history,” Cummings said in an email, speaking generally. “I do not see this changing anytime soon.” 

In 2016, a Knightscope robot deployed at a shopping center in Silicon Valley rolled over a 16-month-old boy's foot. The company apologized for the “freakish accident” at the time, saying in a statement that the robot veered to avoid the child, who ran backwards and directly into the machine. The next year, one flopped into a Washington, D.C. fountain. The company responded with humor, tweeting: “BREAKING NEWS: “I heard humans can take a dip in the water in this heat, but robots cannot. I am sorry.” Li, meanwhile, told the New York Times that the incident was under investigation and a new robot would be delivered free.

When asked about whether contracts remained active, representatives from neither property  responded.

Then there was a trial at the San Antonio International Airport in 2024. Spokesperson Ana Flores said its Knightscope robot struggled with badge scanning, communication, and finding its way around. A door alarm would sound, she said, but the robot wouldn’t respond efficiently. Airport leadership opted to not move forward with a one-year, $21,000 Knightscope lease following a one-month test.

Meanwhile, private security is booming. Security officers have proliferated in recent years as cities embrace private guards, who earn an average of about $40,000 a year, according to the U.S. Bureau of Labor Statistics, about half that of police officers. 

Robotics companies pitch their autonomous robots — some of which can be equipped with cameras, facial recognition capabilities, and license-plate readers — as another set of eyes. They have no family to go home to and are happy to work nights and weekends.

Security robots promise “an alluring form of surveillance as a service,” Andrew Ferguson, a law professor at George Washington University, said in an email to Proof. In some cases their visible video cameras can “deter crime without having to pay a human being a salary,” he said, but it’s not clear that robots prevent or solve crimes better than people.

“They exist as a physical manifestation — a symbol, if you will — that someone is doing something to address potential crime,” Ferguson said. “Robots are a form of security theater, more optics than results.”

CRUMPLED CONTRACTS

Knightscope isn’t the only one seeing interest fizzle.

Officials in Salem, Oregon, powered off their Daxbots after a three-month pilot with the Oregon-based robotics company earlier this year. They had deployed three Daxbot robots to deter trespassing in a downtown parking garage. The robots recorded license plate numbers, captured photos of vehicles violating parking lot rules, and shared data with police.

Complaints about the parkade are “drastically down” since the pilot ended, according to Salem Police Sgt. Trevor Morrison, but it’s unclear if that’s due to traffic-calming devices or “residual results from the Daxbot patrols.” The pilot ended in April, with city spokesperson Nicole Miller citing no additional funding for robot security. 

Daxbots have also disappeared in Tempe, Arizona. A viral video of one blue-eyed bot being outwitted by an onlooker who it demanded “leave the premises” made its way to “America’s Funniest Home Videos.” When asked about the contract, the mixed-use campus, IDEA Tempe, did not respond. Daxbot didn’t respond to Proof’s questions about why the Salem and Tempe contracts ended. 

Adam Pioth, who posted the video, told Proof that he visited the robots frequently. A few months ago, he noticed the robots no longer shouted he was trespassing. Instead, he said, they would, “stop, look at me, and then just keep going really slow.” And then, he said, they were gone.

“Goodbye my friend,” Pioth wrote on Instagram in March, “and thanks for all the fond memories.”

Daxbot is recalibrating its business, rolling into another industry: automated sidewalk assessments. Proof identified at least nine cities nationwide that have contracted with the company since last September to send robots cruising sidewalks to check for compliance with the Americans with Disabilities Act.

Boston Dynamics is best known for its robotic patrol dogs, which were recently deployed to two World Cup stadiums. After Hyundai bought the controlling stake in the robotics company in 2021, it began pushing to develop mass-produced humanoid robots to work in automotive manufacturing.

Atlas, its humanoid robot, delivered the match ball to the referee during a game in July. But the strategy has also spurred rolling strikes at Hyundai Motor’s Korean assembly plants. 

Critics claim police and security robots aren’t proven crime fighters, but are effective at one thing: surveillance.

In 2021, the New York Police Department ended its lease for a Boston Dynamics robotic dog and returned it, The New York Times reported, following public outcry. U.S. Representative Alexandria Ocasio-Cortez called the tool — which could climb stairs and was equipped with cameras, lights, and a two-way communication system — “surveillance ground drones” being “deployed for testing on low-income communities of color.”

Boston Dynamics did not respond to requests for comment. 

“What it’s really doing is building a mass surveillance system that is impossible to hide from,” said Brian Hofer, executive director of the privacy-advocacy nonprofit Secure Justice. “It is not a society I want to live in.”

Steve Rosta, retired highway patrol officer and co-owner of Direct Approach Security Services in Ohio, said robots’ cameras may act as a deterrent, and even catch videos of crimes, but he can’t see a future where the machines are more reliable than a human guard. 

“I’m old school,” Rosta said. “Boots on the ground is always better.”

Satellite imagery has confirmed that Ukraine's Defense Forces destroyed a Russian electronic-warfare system built to jam Starlink at Gelendzhik in southern Russia, the Ukrainian defense outlet Militarnyi reported. Ukrainian drones hit the machine on the grounds of a Russian border-guard base during their sweep of the Black Sea coast. It is one more of these costly jammers Russia has lost, even as Ukraine's long-range drones cannot use Starlink over Russia.

Satellites caught the before and after

Satellite images published by the OSINT community Dnipro Osint show the site cleared at Gelendzhik, in Russia's Krasnodar Krai. The Volna Kupol Garant complex is gone. Researchers called it the third destruction of this jammer recorded in open sources. The strike landed on the grounds of military unit 2156, the Novorossiysk border detachment of Russia's FSB Border Service.

Ukrainian drones likely hit it alongside a 7 August strike on cargo ships near Novorossiysk and Gelendzhik. Geolocated photos from residents point to a repeat strike on the base on 8 August.

Explore further

Ukraine’s newest drones don’t need Starlink. Russia spends $1.5 million to jam it anyway.

The Volna Kupol Garant is built from trailers carrying satellite antennas. It tracks a passing Starlink satellite and beams a strong signal that scrambles the terminals below across up to 20 km². Starlink is the SpaceX satellite-internet network that Ukrainian drones rely on to reach their operators. Each unit costs about $1.8 million. A same-named firm registered in occupied Simferopol builds it. The developer also makes the Volna 3M jammer that Russia has fitted to Mi-8AMTSh helicopters to block drone satellite navigation.

Ukraine has knocked out these jammers before. The 422nd Unmanned Systems Regiment, the Security Service of Ukraine's (SBU) Alpha special-operations center, and other Ukrainian units have struck such systems, including in occupied Crimea. The complex's weakness is its own power: beaming enough energy to reach a satellite makes it loud and easy to find. 

Explore further

Russians deploy massive $1.5M Starlink jammers, Ukrainians are blowing them up

And deep inside Russia, there is little Ukrainian Starlink to jam — SpaceX geofences the network to Ukrainian territory, and Elon Musk has refused Kyiv's push to use it for strikes into Russia so far.

The wider hunt

Ukraine's drone crews have spent 2026 tearing into Russia's air defenses and rear. The same run of Black Sea strikes that wrecked this jammer also burned a Russian S-400 air-defense system on the Gelendzhik coast. 

SAP, one of the world’s biggest software companies, suspended most travel and hiring last month because of AI’s soaring cost, with exceptions only being made for AI-related travel or hires, according to an internal SAP email obtained by 404 Media. 

Bloomberg reported on the email and the freezes in July, but a current SAP employee said the bans are still in effect. The employee said the company had a global employee meeting recently where this was brought up again. They added SAP is currently rolling out a newly created AI tool to the entire company “which I can only imagine massively increases the costs.” 404 Media granted the source anonymity to protect them from retaliation.

The email highlights how companies both big and small are coming to grips with the reality of AI’s cost. Rather than being a massive cost saver, in some cases companies are throttling employees’ AI use as it spirals out of control, and scrambling to find other ways to keep costs down.

Microsoft has introduced new limits to how much its engineers can spend on AI tools at work and told employees that maximizing AI use internally is not the company’s goal. 

This makes Microsoft one of the last major companies to rein in its employees’ expensive AI use. Scaling back maximalist AI use, or what some companies have called “tokenmaxxing,” is a trend we’ve covered in recent months as the price for using AI has increased while not always delivering commensurate productivity gains. 

Ukrainian aerial- and sea-drone attacks have turned Russia's busiest seaport into a place cargo ships increasingly refuse to sail, choking both imports from China and exports of grain and oil, according to The Moscow Times. Logistics and trade executives describe a route to Novorossiysk grown too risky to use, and a Russian grain-industry group warns the Black Sea export corridors could shut down entirely. The strikes that once spared container ships have now reached them too.

The China route importers are told to skip

"On the China–Novorossiysk route everything is now difficult, the delays are huge," one importer told The Moscow Times. 

Freight forwarders now tell clients to avoid the route outright, three colleagues said. Aerial drones and unmanned boats make the approaches too unpredictable. Earlier strikes fell mostly on grain carriers and oil tankers. Ukrainian drones hit a container ship carrying imports in late July. Then, overnight on 1 August, a sea drone sank the FESCO vessel Yanina about 130 miles offshore. The crew got out. The cargo did not.

Explore further

Ukraine hits 201 Russian shadow fleet ships in three weeks. Commander says Crimea’s supply route is now blocked indefinitely

Sailings drop as the campaign widens

Large-ship sailings from Novorossiysk fell 38% in July, the Moscow Times says, citing ship-tracking data, with tanker runs down by half. 

"But now the situation will get worse, because container ships have also become a target," a logistics manager told the publication. 

This is one edge of a wider Ukrainian drone campaign against Russian shipping, codenamed MoLoChKa. The name is an acronym for "Moscow will fall through Crimea." Its own public scoreboard counts 206 vessels hit in July by the Unmanned Systems Forces (SBS) alone, with another logged in August. The SBS have mostly burned the small tankers and support ships that keep Russia's sanctioned shadow fleet moving.

Explore further

Ukrainian drones slash Kerch ferry crossing capacity by 75%, as daily strikes continue on Russian vessels and occupation energy infrastructure

What Russia stands to lose

Novorossiysk carries every fifth ton of Russia's seaborne exports: Urals oil, Kazakh crude, wheat for Egypt, and fertilizer for Asia and Africa. 

Attacks across the Azov–Black Sea region could cost Russia up to 30–35 million tons of wheat exports this season, a grain-exporters' union estimated. That would gut a harvest analysts expected to top 44-45 million tons. 

"Systematic attacks that began in July could soon lead to the complete blocking of the export corridors of the Black Sea basin," the union said. 

Ukraine's drones had already cut into Russia's seaborne grain trade this summer, and repeated hits on the port's main oil terminal have dented crude loadings. Russia keeps pouring money into the port, building a new terminal worth 120.7 billion rubles (about $1.5 billion), even as ships stay away, The Moscow Times said.

A drone came down on a beach full of Russian holidaymakers, and the evidence points to Russia's own air defense bringing it there. On 3 August, drones attacked Krasnodar Krai, and one crashed onto a beach in Arkhipo-Osipovka near Gelendzhik, killing at least six people, three of them children, and injuring around 40, Krasnodar Governor Veniamin Kondratyev said.

Russian outlet Agentstvo and OSINT analysts assessed that the drone likely fell because of Russian air-defense fire rather than a direct hit, citing gunfire audible in footage before the drone dived and an abrupt change in its flight path.

The beach sits directly in front of the Rassvet resort, a facility of NPO Mashinostroyeniya, one of Russia's key missile-manufacturing enterprises.

The warning system did not sound

Survivors say they had no notice. Local residents and holidaymakers complained that no air-raid alert sounded at all, though warning systems are installed in the village, and the Russian channel Astra also noted that no alert was heard in the region during the attack, per the reports.

People on the beach were in the open, with little chance to take cover, even if they had been warned. The absence of a siren in a resort packed at the height of the summer season left sunbathers and swimmers with no warning that a drone was overhead and being fired at.

The target was a missile maker's resort

NPO Mashinostroyeniya develops cruise missiles for the Russian army, and its Rassvet recreation base was the apparent target roughly 300 meters from where the drone came down, the Supernova Telegram channel reported.

Ukraine strikes Russian military-industrial sites across the country, and a missile-development enterprise is a legitimate military target. What turned a strike near a defense plant into a mass-casualty event on a beach was, by the analysts' account, Russian air defense knocking the drone down over the crowd rather than away from it. Ukraine had not commented on the strike.

The dynamic is one Ukraine knows intimately in reverse

Russian air-defense missiles and the debris of intercepted drones regularly fall on Ukrainian cities, killing civilians, and Ukraine has long documented how interception over populated areas kills the people it is meant to protect.

404 Media has obtained a coaching guide that Flock surveillance gives to police about “how to speak to city councils about public safety technology.” The handbook highlights how Flock and police team up to convince cities to buy and keep its automated license plate reader technology, even when there is widespread public opposition to it, and encourages police to “own the narrative before someone else does” by championing the technology before citizens can oppose it during public comment periods.

The PDF guide notes that the general public and cities now “increasingly expect transparency, oversight, and accountability alongside public safety outcomes,” and tells police to not argue with people who believe that Flock’s license plate readers are “mass surveillance.”

“One of the most common questions agencies hear today is whether license plate recognition (LPR) technology constitutes mass surveillance. Many leaders instinctively respond by attempting to refute the claim. Flock's Jamie Hudson recommends a different approach,” the guide reads.

“Don’t avoid the concept of mass surveillance because you’re not going to convince opponents that it’s not,” the company recommends. Flock tells law enforcement agencies they need to try to convince city council and city managers that the technology is worthwhile before meetings with the public occur; that they need to have a “carefully scripted presentation” ready to go; and that police need to say they want Flock because they want to keep the community safe: “You care about your community. That’s why you’re bringing this in.” 

Flock began offering this guide as part of a broader attempt to coach police on how to push back against criticism of its policies and security practices, many of which 404 Media has investigated and shed light on. These include the fact that Flock data was regularly making its way to Immigrations and Customs Enforcement (ICE), often in violation of sanctuary city and state laws; that Flock was used to search 83,000 cameras nationwide for a woman who had an abortion in Texas; and that Flock has been used by police to stalk people and surveil protesters. These investigations and broader concern over the surveillance state have led many cities to hold city council meetings to reconsider their Flock contracts, and this Flock-produced guide is an attempt to help police shape the narrative in a way that will either convince cities to buy Flock or to keep their contracts. 

The guide is associated with a Flock webinar for police called “How to Speak to City Councils: Meeting the Moment with Confidence,” which included modules on “how to address misinformation with clarity.” After public opposition late last year, the company began to offer Q&A sessions with its CEO, Garrett Langley, for city council members, police, and mayors to address what Flock described as an “era of unprecedented misinformation.” 

“The recent headlines about our company are largely a result of this environment,” the company told cities. 

“Opponents have a very carefully scripted narrative. They come prepared. You should also have a carefully scripted presentation that addresses those concerns ahead of time,” the guide says. “The agencies that navigate these conversations successfully rarely wait until a council meeting to educate stakeholders. They brief city managers early. They meet with council members before votes occur. They share policies proactively and answer questions before public comment periods become the first introduction to the program.” 

The guide also tells police that they can convince city councils that Flock is worth the monetary cost by conveniently not focusing on how much the cameras cost, but by “reframing the financial discussion itself” to focus on “the cost of unresolved crime.” Flock also writes that much of the opposition to its technology is happening because people “do not understand how it works or how it is governed.” This idea is one that has been regularly repeated by Langley over the last several months. 

Surveillance companies regularly try to get police to act as quasi salespeople and spokespeople for their companies, pitting a private company and taxpayer-funded law enforcement on one side and citizens on the other. “For years, surveillance vendors like Flock Safety have shaped policy debates cities are supposed to run independently — staging council ‘prep calls’ and exploiting a basic asymmetry: the vendor controls the facts, and city staff are rarely positioned to challenge them,” Sarah T. Hamid, director of strategic campaigns at the Electronic Frontier Foundation told 404 Media after reviewing the guide. “The financial interest is obvious. Flock isn’t just selling surveillance, it’s scripting the public case for buying it. Because Flock treats public trust as a messaging problem rather than a governance outcome, that script keeps officials focused on ‘accountability’ in the abstract instead of the concrete harms and documented abuses its network has already enabled.”

404 Media has watched numerous city council meetings around the country where police talk about how Flock is a critical law enforcement system for them; in many cases, a police chief will speak about Flock and then introduce a Flock employee to give a presentation about the surveillance system. On Monday, 404 Media published an interview with a former Flock government affairs manager who regularly pitched the technology to cities at public meetings. An activist who has been pushing back against Flock in their community and who shared the guide document with 404 Media said that they have regularly seen the strategies suggested by Flock deployed in city council meetings they have attended and watched. 404 Media agreed to keep the activist anonymous to protect them from retaliation.

“I think this document shows a coordinated effort from Flock Safety to compel law enforcement agencies to convince our elected leaders to represent their interests as a company rather than the interests of concerned citizens,” they said. “I have watched many meetings locally in my city and my state and across the country, and you can see the techniques used in this ebook in the presentations given by law enforcement. Pivoting conversations away from concerns about mass surveillance and directing them towards procedure and governance is a vehicle that's used to downplay the concerns of privacy-minded citizens. We have every right to expect our elected leaders to listen to us, and it's very common to see city councils vote with a supermajority in favor of approving Flock contracts despite standing-room only attendance at city council meetings with little to no public support for this product.”

The guide specifically highlights several supposed success stories in which communities had very real concerns about Flock but ultimately decided not to get rid of the technology. For example, it highlights how Flock was able to get a vote in favor of its technology in Oakland, California, despite it being “one of the most scrutinized public safety technology debates in the state,” with “more than 140 public comments” and opposition from the city’s Privacy Advisory Commission: “The conversation shifted when officials stopped asking the public to trust the technology and started showing how the technology could be audited, reviewed, and held accountable.” 

It also tells the story of Richmond, California, which allowed its Flock contract to temporarily lapse after the city’s cameras were included in the company’s national lookup tool. City officials there worried that their cameras’ data would be accessed by ICE, in violation of California and local law. “After concerns emerged around data sharing and sanctuary city policies, the city's program was paused and subjected to intense public scrutiny,” the Flock guide says. “Rather than relying on generalized claims about effectiveness, department leadership presented two and a half years of local results, including 274 arrests and 259 vehicle recoveries connected to the program. The council ultimately voted 4-3 to reinstate the system.”

Flock did not immediately respond to a request for comment. 

Ukraine set a major Russian oil terminal ablaze on the Black Sea coast. Operators of the SBS 413th "Raid" Regiment, in a joint operation with comrades from the 1st Separate SBS Center and the Security Service of Ukraine (SBU), struck the Taman maritime terminal near the village of Volna in Krasnodar Krai on 30 July, the regiment reported.

The strike destroyed a volume of fuel equal to 1,000 rail tank cars. According to updated data, 12 reservoirs, each holding 5,000 tons, were destroyed at the Taman terminal, and the fire covered more than 17,000 square meters.

The terminal is a key southern export node

Taman is one of the most important fuel facilities in southern Russia. The terminal specializes in the transshipment and export of oil, fuel oil, diesel, and liquefied petroleum gases, operates a large tank farm, and supplies fuel to Russian forces, according to the regiment.

Destroying 12 of its tanks strikes both Russia's fuel exports and its military supply at once, the dual purpose of Ukraine's campaign against Russian oil infrastructure, which targets the refineries, depots, and terminals that turn crude into export revenue and military fuel.

Satellites confirmed the fire

The strike's effect was visible from orbit. The Ukrainian drone strikes caused a large-scale fire on the terminal's territory, recorded by the NASA FIRMS satellite fire-monitoring service.

The reported date corroborates that something large burned there, moving the claim beyond a single belligerent's word.

The unit pioneered Ukraine's long-range drone strikes

The 413th "Raid" Regiment is one of the units that started Ukraine's deep-strike campaign. In 2022, at the start of the full-scale war, a group of special operations fighters first used drone technology to reconnoiter and strike enemy forces at distances up to 500 kilometers, the origin of the regiment.

The Taman strike extends the near-nightly Ukrainian campaign that has idled a large share of Russian refining capacity and now reaches the export terminals on the coast, raising the cost of the oil trade that funds Russia's war.

On Thursday, Google introduced a new AI feature into Google Earth which lets anyone fabricate all sorts of misleading or straight up inaccurate satellite imagery, from making it look like a specific place has suffered a drone strike to manifesting a nuclear plant in Iran.

Usually, Google Earth is an exceptionally useful tool for open source intelligence (OSINT) analysts to digitally monitor areas of interest and see how they change over time, say, during a conflict or disaster. Now, Google Earth can easily be used as a tool for disinformation. 

Following 404 Media’s reporting that book database company ISBNdb claimed to source printed books to then sell to AI companies for AI training, the company deleted the part of its website offering the service and walked back claims that it would train AI models, and instead called it “a test of market interest.” 

On July 30, nine days after 404 Media’s reporting, ISBNdb added a note to its homepage and an update on its news page about the change. “We've seen the recent coverage about a marketing landing page on our site, and we understand the concern it raised. The facts: ISBNdb has never purchased, scanned, or sold a book — for AI training or anything else,” ISBNdb wrote. “We don't train AI models, and we never have. The page was a test of market interest; no such service was ever brought to life. We've taken the page down. Our job is helping people find books. For more than two decades, ISBNdb has been the card catalog of the book world — the data behind how bookstores, libraries, and reading apps connect readers with titles. Data about books, not the books themselves. That hasn't changed.” 

In February, a content creator from New Zealand named Harry Chang posted a YouTube video called “This AI TikTok Shop Video Made Me $67,420 (Here’s How).” In the video, Chang and another YouTuber, Jimmy Farley, describe how Chang created a viral marketing video for a supplement company called Rosabella called the “Nigerian SECRET to CLEAN LIVER!!”

Chang explains that he copy-pasted the script from a video posted by an account called “liverboosthub11” and tweaked it to suggest the supplement he was marketing is “something that’s been used in Asia or Africa for a long time that a lot of people don’t know about in America.”. In Google’s VEO 3, he created an AI-generated Black woman wearing a surgical mask in the foreground of the video, pointing up at another AI-generated Black woman (created in a tool called HeyGen) wearing a pink dress and standing on a stage. He directed the woman in the foreground to have a “strong African American accent,” and to say, “Why is nobody talking about what this hoe said?! If you’ve got issues with that belly, must watch!” 

“Builds curiosity, builds intrigue,” Chang says about his creation. “People want to know, ‘damn, what did she say? What did that ho say?’” To make the AI-generated woman on stage read the script he took from liverboosthub11, he pulls up the popular AI voice generator ElevenLabs. He scrolls through a list of voices that he had created, including “African American Woman Organic,” “Black Man 1,” “ORGANIC White Southern Woman,” and “Sad Black Woman in Car.” He settles on a voice called “Latisha 1.” He syncs the voice with the AI-generated videos he created in the video editing software CapCut. This AI video goes on to get 1.3 million views on TikTok and apparently earned him tens of thousands of dollars in affiliate sales.

In another video, Chang explains how he has made tens of thousands of dollars using AI influencers. “I’ve even had my clients buy me Rolexes for selling so much of their products,” he says. 

The video’s title is “How I print $51,000/month profit with AI influencers (feels illegal).” 

A new lawsuit argues that the strategy is, indeed, illegal. (After 404 Media asked for comment for this story, several of the YouTube videos mentioned in this article were deleted).

In February, a supplement company called Humann sued a competitor called Ambrosia Brands because Ambrosia, through the supplement company called Rosabella, allegedly directed and influenced the creation of hundreds of TikTok Shop videos and ads featuring AI-generated “doctors” that oversold the supposed benefits of Rosabella’s products. Rosabella’s AI marketing practices have previously been written about by 404 Media and The New York Times, but were most thoroughly explored in an excellent episode of the podcast Conspirituality. 

The lawsuit reveals new details about how this group of 20-something YouTubers built their army of AI-generated influencers. In practice, Rosabella is more of a social media AI content hustle and AI marketing exercise than a supplement company. What happened in this case is the same type of spam and buy-my-course to get-rich-quick strategy that we have repeatedly written about, only this time the slop is being used to shill supplements largely to the elderly. Rather than payouts coming from the number of views a video gets on social media, the payouts are commissions on products sold. The lawsuit was spotted by the lawyer Rob Freund on X.

“All these guys are ex-dropshipping guys,” Mallory DeMille, who studies the wellness and supplements industry and who reported the episode of Conspirituality, told 404 Media. “They could have chosen anything to sell to make AI content out of, but they chose supplements, and it’s interesting they chose supplements because it’s such an unregulated market where [they] can basically pump out whatever product they wanted to with very little oversight. On the marketing side, it’s also pretty unregulated and there’s a lot of real [human] influencers making unfounded health claims without there being many consequences. In terms of ease of making money — wellness, they chose this industry for a reason. I think it’s pretty seamless, has proven to be seamless and now they’ve sold a fuckton because of how easy it is.”

The lawsuit highlights a series of TikTok videos—like the ones I described above, and most of which are still online—featuring AI-generated doctors, TED Talk-style speakers, and videos that are essentially identical to the ones Chang has repeatedly taught people on YouTube how to make. And hundreds of additional videos promoting Rosabella that are not highlighted in the lawsuit are trivial to find on TikTok. Many of them have hundreds of thousands or millions of views and seem to make wild promises about what Rosabella supplements can do.

In another video, Chang explains why his favorite products to sell fall into the “elderly health” category: “It’s an extremely profitable niche, especially in the U.S., guys.” 

“In the U.S., they have very high demand for health products,” he says. “They don’t have any free healthcare, right? People over 35 literally are very concerned about their health […] You have moms who buy 100 supplements and put them all in their cupboard. It is crazy. It’s crazy.” 

“When you have such a high, problem-solving niche, it’s very easy to write scripts for, very easy to innovate, and just continuously make money,” he says, adding that he usually generates AI influencers who themselves look old.

“They’re relatable and credible,” he says of three older-looking AI people he shows on screen, one of which is shilling beetroot powder from Rosabella. “Why? Because they are old. People are more willing to listen to old people only because they see them as more wise and intelligent […] and look how credible they look. They’re speaking on stage like TED Talk-type style and they’re more relatable because they’re more similar to the age we want to target. If you’re trying to sell health products to a 50-year-old, well, make your avatar 50 years old.”

On TikTok, Rosabella used a mix of paid influencers and AI-generated characters to make unfounded medical claims about its products and about beetroot more generally, the lawsuit alleges. “In several TikTok Posts, the post purports to show a doctor, medical professional, or other medical authority espousing the health benefits of Defendant’s products. All or nearly all of the ‘doctors’ featured in the TikTok Posts are AI-generated and fictitious, making the claims in the advertisement false and/or misleading,” the lawsuit notes. “For example, in a June 14, 2025 TikTok post, influencer ‘poormaninla’ purports to depict a doctor in a whitecoat that promotes the alleged health benefits of Defendant’s products. Before the ‘doctor’ begins speaking, a separate speaker is imposed on the screen in a surgeon’s or nurse’s scrubs.”

The “poormaninla” account is still up on TikTok and its videos are almost entirely AI-generated. Several of the videos have hundreds of thousands of views. “The best food for Black women to eat if they want a slim stomach is not turmeric, it’s not ginger, and it’s definitely not blueberries. Just one teaspoon of this food reduces gut inflammation,” an AI-generated man in a lab coat says in one.

You can see some of the videos promoting Rosabella here:

The lawsuit argues that Rosabella is “orchestrating a misinformation campaign on TikTok through its network of influencers,” who “make various misrepresentations about the health and wellness benefits of Defendant’s beetroot products, which are entirely unfounded.” The lawsuit alleges that Rosabella, through a private Discord channel, offered extensive coaching to content creators on how to make AI-generated TikTok ads; that many of these ads featured fake doctors and other AI-generated people who were made to look authoritative; that some of these videos were racist; and that when people bought Rosabella products through TikTok Shop links on those videos, the creator of those videos would earn a commission. Humann’s argument is that Rosabella’s “false and misleading representations undermine public confidence in other beet products, like Humann’s SUPERBEETS products.” 

Ambrosia says, essentially, there is no evidence it told people what to do in the Discord channel. In court filings, Ambrosia claims that Humann “fails to allege any facts that plausibly show Ambrosia induced or materially contributed to the third-party conduct it complains of.” It remains unclear whether what Rosabella was doing is illegal in the unregulated world of supplements in the U.S., or whether a competitor could win a false advertising lawsuit like this. But there is no doubt about Rosabella’s strategy, what their motivations are, and the incredibly close connection between Rosabella and the network of content creators who made, conservatively, hundreds of AI-generated videos.   

The lawsuit highlights how common fully AI-generated marketing has become on platforms like Instagram and TikTok, and nods at, but does not dive into, the complicated web of YouTube hustlebros that have largely given rise to this practice, and the social media platforms that have incentivized and promoted AI-generated spam and fly-by-night supplement companies. Rosabella is a company that has already been subject to an “extensively drug-resistant salmonella” recall by the Food and Drug Administration. But it is maybe better understood not so much as a supplement company but more as a branding exercise and vertical video content hustle by the same types of AI spammers and buy-my-course bros we’ve written about numerous times over the last few years.

In a video about “why you need to be working with Rosabella” qposted by an account called “Luca Washenko” on the online course sales platform Whop, a man brags “we paid out over $400,000 last month to creators.” He says that individual people have made more than $300,000, and shows “proof and dates of our creators getting tons of views consistently. Got 55 million right there […] I am one of the lead coaches in the Rosabella server.” 

But Washenko isn’t just a creator helping to advertise Rosabella. He is the company’s founder. A YouTube video repeatedly alludes to this, and mentions how much money he and his army of affiliate creators have made selling Rosabella products on TikTok Shop. Washenko is also listed as the cofounder of Rosabella on several trademark filings I found, and his LinkedIn lists him as the founder of “MNY Ventures,” a company that has the Rosabella logo on LinkedIn. Clicking through “MNY Ventures” goes to a LinkedIn page for “Rosabella,” which has several job listings for AI video editors: “MNY Ventures is home to one of the fastest-growing supplement brands in the world, built on the back of a high-performance, results-obsessed culture. We don't just create ads; we create market-leading campaigns that generate massive revenue,” one of the LinkedIn job listings reads. “Your mission is to lead the production of our high-converting AI videos quickly and at high quality. You will be responsible for consistently creating on-brand and compliant video content based on proven formulas designed to maximize reach, ensuring MNY Ventures maintains its position as the #1 leader in AI video marketing for e-commerce.” The listing adds the person will need to produce “10 high-quality AI videos per day, following our preset scripts and styles.” 

In a video called “Inside a TikTok Shop Meetup with Million Dollar Creators,” there is no doubt about Rosabella’s strategy, Washenko’s motivations, or Rosabella’s close relationships with the people spamming AI-generated content shilling its products. Rosabella is just the latest of Washenko’s creations. His previous claim-to-fame was selling caffeine vapes on TikTok. 

“There have been creators that are now millionaires from working with Rosabella,” Washenko says in the meetup video. “It’s one thing if you make money, it’s another thing if you can help everybody around you make money. An event like this where you can go face-to-face and shake hands with people and they say, ‘You’ve changed my life.’ And I’m like ‘You’ve changed mine.’ Those are the moments that are so special, especially when everything is remote. It’s online. It’s Discord, that’s one thing. You look them in the eye and they say, ‘I was able to pay my mortgage when I lost my job because I was making videos for you.’ That’s a different feeling.” 

Washenko explains there are two reasons why people should make content promoting Rosabella: “If you want to be the Tiger Woods of creator, if you want to be the Caitlin Clark, you want to be the Michael Phelps, you want the golden rings you want to be the greatest, to be the number one, that’s what we’re all about. That’s the Rosabella family,” he says. “Along the way, you’ll make a ton of money.” 

Later in the video, Washenko stands on stage addressing a crowd, shouting out “all the people I’ve been talking to on Discord from day one, from 18 months, I started Rosabella in my mom’s basement.” He shouts out all the creators who he’s helped make rich and who helped make him rich. He then says there’s one other “person I want to shout out today.” He calls up Harry Chang, the YouTuber who made the video “How I print $51,000/month profit with AI influencers (feels illegal).” Washenko presents him with a Rolex. “I cannot be any more proud to be working on a brand with you,” and for that, I want to give you this watch today. It’s a Rolex, by the way.” 

After 404 Media asked for comment on the video, it was deleted from YouTube.

DeMille, who reported the episode of Conspirituality diving into Rosabella, Washenko, and Chang, told me the men are “just actively bragging about it online. I can’t believe they’re openly talking about it like this.”

“A lot of the videos of these AI slopfluencers, they’re using narratives that real-life wellness influencers have seen success using, but they’re taking these narratives and inputting them into whatever AI systems they’re using and making people look whatever age they want,” DeMille said. “There’s no thought behind it. There’s no anything behind the content that they’re turning out. These videos they’re replicating in their AI machines are also being used to sell something, but it’s being generated into something else to sell something else. It’s this inception of bad information by people who don’t actually care.”

“These guys behind these AI slopfluencers have proven that they don’t care, and they obviously care more about money and wealth than they do about health,” she added.

Washenko did not respond to a request for comment. Ambrosia Brands did not respond to a request for comment. 

A request sent to Rosabella was returned by someone named Emmanuel Obonga. The response said, “We don’t currently have an active affiliate program. However, we’d be glad to keep your information on file and reach out if we revisit or launch one in the future.” In the course of reporting an earlier story about the AI influencer company Doublespeed, Rosabella previously told 404 Media that it “does not use Doublespeed or any AI-generated accounts to promote our products on TikTok or any other platform. We are committed to authentic engagement and building genuine connections with our community. Regarding the claim about being viral on TikTok, this is based on organic content created by real customers and creators who love our product. We’ve seen a lot of positive buzz and user-generated videos that have helped spread the word naturally.”

Daniel Martens, a lawyer representing Humann, told 404 Media Rosabella’s strategy “makes [the] whole supplement industry look bad. If you’re masquerading as a doctor telling consumers how promising all these benefits that don’t exist [are], that’s harmful.

LinkedIn, a social network awash with long AI-generated posts from executives and other corporate workers, has introduced a new button that users can click to flag if a post “seems like AI slop,” according to 404 Media’s own tests.

If you have been anywhere near LinkedIn in the past couple of years, you have undoubtedly seen users posting blatantly AI-generated missives. Often these posts take some sort of news event, and opine on how this relates to thought leadership, or some other LinkedIn brainrot term. It’s also pretty wild the button specifically uses the term “AI slop” and not, say, “It seems this was generated with AI.”

On Wednesday the power flickered in homes from Washington DC to Chicago. The cause was a data center disconnecting from PJM — a massive power grid that connects 13 states including North Carolina, Virginia, New Jersey, and Michigan, a span that covers 67 million customers. The incident points to a larger problem with the compute warehouses fueling the AI boom: It’s physically and economically impossible to build power infrastructure fast enough to meet the demands of AI data centers and when problems occur, everyone will pay the price.

AI requires massive amounts of computer hardware to scale. That hardware is housed in data centers, and data centers demand shocking amounts of electricity. More data centers mean more demand on the grid which raises the costs of energy for everyone. This is one of the reasons people hate data centers and have begun to fight their construction.

To bypass the issue of grid demand, many data center builders have promised to power the buildings themselves. The problem with that plan is that it’s easy to build a warehouse full of GPUs. Building power plants and transmission lines to power those warehouses will take years and cost billions more than the data centers. A skilled and efficient builder can complete a data center construction in under a year. Building new energy generation to meet that data center's power needs could take a decade. 

Substack contributors are rejecting the company’s decision to flag AI generated writing on the newsletter publishing platform, with several creators saying the new AI detection features are a “witch hunt.” Some Substack users who use AI to help them write or edit articles say the new policy unfairly discriminates against them, while other users who don’t use AI in their writing at all are worried that Substack will mistakenly flag their writing as being AI generated and tarnish their reputation. 

“I’m not going to apologize for using AI in the creation process. I wrote for 20 years without AI, I could do it again if I wanted to,” Mack Collier, who has a small Substack called Backstage Pass, wrote. “My output would fall, my posts would be less structured, and it would be more obvious that they needed a good editor. Using AI improves the overall quality of my writing. That’s why I use it.”

“These detectors are notoriously, wildly inaccurate,” Alice Lemee, a ghostwriter and digital writing coach, said in a video she made about Substack’s new AI detection features. “All it takes is one false accusation for a writer to have their reputation almost irreversibly tarnished.”

Substack announced the AI detection features on Tuesday in a post from CEO Christ Best, who said that “It’s getting harder to tell what’s real on the internet” and that “when content made by no one takes over parts of the internet that are supposed to be human, it pollutes the commons and makes it hard to discover and hear human voices.”

Best explained that Substack has partnered with the AI detection tool Pangram, which is now built in to Substack and allows any user to scan a piece of writing. Pangram then produces a percentage-based score determining how much of the writing was AI or human generated. (Disclosure: Pangram previously bought an ad on 404 Media). We’ve covered research from Pangram, or research that relies on its AI detector before, showing how AI generated writing is flooding every corner of the internet. 

As we’ve previously reported, AI detectors are not perfect, and Pangram itself is not immune to labeling human content as being AI, and labeling AI as human. Max Spero, the CEO of Pangram, recently told us that the company is constantly working on minimizing errors, and that it estimates its false positive rate at roughly one in 10,000.  

“We have now built the mirror image. Pangram is a real machine, trained on vast amounts of text to make a probabilistic guess, and we have pointed it at your writing to work out whether a person is hidden inside,” Sam Illingworth, a professor and the author of Slow AI, said on his Substack in a post titled “Substack’s AI Detector and the Return of the Witch Hunt.” “To decide if there is a human on the other end, Substack asks a machine.”

Substack also added a way for users to add a “How I make this” statement, where they can explain their writing process and disclose if they use AI. 

“We’re not against people using AI to assist their work, and we think people should be free to choose which tools they use to express themselves,” Best said. “Some on the platform are human-writing maximalists, others are publicly exploring the frontier of using AI tools while sweating the details to make work they stand behind. We use AI all the time at Substack to write software, do research, and build product features like clipping, translations, and more. But people should know what they’re getting.”

“I think the best way for Substack to address this is to invite further dialogue,” Illingworth told me. “I actually really like the new feature that lets authors tell their readers how they construct their newsletters using AI, and this is the kind of dialogue they should encourage rather than ascribing a score that we know to be incorrect. My worry is that not only do these AI detectors generate false flags for non-native English speakers and neurodiverse writers, but they also remove any opportunity for dialogue because they start from a position of suspicion. Whereas what we need to be doing is developing opportunities for trust between readers and authors.”

"The tools we've introduced today are intended to increase transparency and give readers more context, not to prohibit or penalize AI-assisted writing. They do not impact discovery on the platform,” a Substack spokesperson told me. “We're also encouraging Substack publishers to add a ‘How I make this’ statement, where they can explain their process directly, including how they use (or don't use) AI, and set expectations for readers. Creators can disable detection on their posts pre- and post-publication, as well as report and remove scans on their own work that they believe are mistaken. Find more information on how the features work here.”

Substack’s attempt to detect and label AI generated content was also celebrated by many readers who are exhausted by AI slop and other internet platforms that can’t or don’t want to label it. I just published story about how there is so much unlabeled AI slop on Spotify that people are now creating their own sites to detect and label it without the company’s help.  

But the backlash from some Substack users shows that companies can’t just flip a switch if they want to ban or give users the ability to filter out AI content from their feeds. As the flood of AI writing and images we see online and the real world every day makes clear, generating slop is easy. What we’re going to do about it is still an unsolved problem.

Slime Dot, a young R&B artist from Las Vegas, currently has over 100,000 monthly listeners on Spotify. In an interview with GQ in May, after sharing an image of herself posing with Drake on Instagram, Slime Dot directly denied accusations that she was an “AI artist,” saying “The truth doesn’t matter these days. People are always gonna try to explain something they don’t fully understand and believe what they want.” GQ originally credulously published the interview and said it was “debunking the AI rumors.” Eventually, GQ added a note to the end of the article conceding Slime Dot was an AI avatar, but said the “talent behind the music remains undeniable.”

This confusion could have been avoided if Spotify had done what many of its users and critics have been begging it to do since AI music generators like Suno and Udio made it trivial to flood the internet with AI music. Spotify could flag Slime Dot and other AI-generated music on its platform as such. Other platforms, like YouTube, already do this. 

Since Spotify doesn’t do that, users can visit SoullessMusic.com, a database for “AI artists hiding on Spotify. No bands, no studios, no soul, just machines and melody.” There, Slime Dot is listed as “almost certainly AI” based on an analysis of three of the tracks and the fact that Slime Dot released such a large number of tracks in a short period of time. Alternatively, users could go to SlopTracker, where they can upload files or copy/paste links to Spotify tracks to run them through a tool which will tell them, with varying degrees of confidence, whether a song was AI-generated or not. SlopTracker found that Slime Dot’s track “Fully” was 95 percent likely to be AI generated by Suno. 

“I didn't know what's AI and what's not,” Graeme Fulton, who created SoullessMusic, told me. “And then I'm on Instagram Reels and I see loads of artists unhappy saying some artist has popped up with their song and just copied it. And there's loads of different cases where they [AI artists] would copy an artist's look and appearance […] sort of just stealing from artists.”

Last year, I wrote a story about an exceptionally bad example of what Fulton is talking about. A scammer used Spotify to publish AI-generated music under the name of a real, dead musician, seemingly in an attempt to syphon money from his popularity via streams. After I published that story I heard from several (living) artists who said the same thing happened to them. A similar case involving a Danish jazz musician was covered by Danish publications in April. 

Another reader recently told me about a popular Spotify artist that was publishing hours of Iranian jazz every week. The channel, Qajar Jazz, which popped up in December of last year, has 29,579 monthly listeners, gained more popularity after the start of the U.S. war with Iran and the channel making vague allusions to resilience and preserving culture in video descriptions and comments on YouTube. Unlike Spotify, Qajar Jazz videos on YouTube are labeled as being AI-generated in the fine print, though it’s unclear if the channel tagged itself as such or if YouTube did. Both SoullessMusic and SlopTracker detect it as being AI generated. 

Qajar Jazz did not respond to a request for comment. 

Many of the YouTube comments on Qajar Jazz videos indicate that people think the music is made by real people. In addition to creating a database of AI generated music on Spotify, SoullessMusic also collects social media posts from people who say they were similarly tricked by AI generated music. Some posts are from musicians who lament the fact that they are now competing with algorithms that can produce thousands of tracks with no effort, and others are from artists whose music was directly ripped off by AI artists. 

View this post on Instagram

A post shared by 𝓘𝓼𝓪𝓫𝓮𝓵 (@isabel_music_official)

View this post on Instagram

A post shared by UMAW (@weareumaw)

Fulton said he started looking around on GitHub for open source AI music detection tools, and ended up building a new, open source detector that relies on open source models called SONICS, the lofcz vocoder fakeprint detector, and other open data sets. SoullessMusic also  scans a track’s metadata for any tags that might tie it back to a specific AI music generator, and Fulton says he also has a script that scans Wikipedia for existing entries on an artist. Users can submit tracks from Spotify to the site they suspect are AI generated, and Fulton reviews them manually before adding them to the database. Despite all of this, he admits the system is not perfect. 

Both Soulless Music and SlopTracker use AI and other automated tools to detect AI music. As we’ve written previously, using AI detectors to detect AI content is an inherently flawed process that can lead to false positives. When I asked him how he handles cases of false positives, Fulton said “it's not ideal when that happens, but it's going to happen because it's still quite tricky to check what's AI and what’s not.”

Spotify’s lack of transparency and unwillingness to label AI-generated music makes it impossible to track how much money its making on the platform. Last year, the company announced that it would it would AI disclosures on AI generated music, but I haven’t seen it on any of the dozens of tracks I reviewed for this story. Udio and Sudo both include inaudible digital watermarking in their generation, which in theory could make it easy for Spotify to tag these tracks. 

"We're employing a layered approach that combines enforcement, artist controls and greater transparency," a Spotify spokesperson told me after this article was first published. "Over the past year, we've introduced policies targeting harmful AI-related behaviors like spam, impersonation and deceptive content, alongside new tools that give artists more control and listeners more context. These include Verified by Spotify, which helps listeners identify authentic artist profiles; AI Credits, where artists disclose when and how AI was used in creating their music, with tens of thousands of AI credit disclosures now being submitted each day; and Artist Profile Protection, which gives artists more control over what appears on their profiles. We are continuing to build on these efforts.”

Spotify also pointed me to several tracks it had flagged as AI on the mobile app, like Better Times by Mike Mana.

Soulless music attempts to quantify that number based on publicly available streaming data. The small number of AI artists tracked on SoullessMusic generate an estimated $5.7 million a year, with the most popular AI artist, mikeeysmind, generating $1.5 million annually alone. In April, Deezer, a Spotify competitor that attempts to tag all AI-generated music on its platform, said 44 percent of all new music uploaded to its platform is now AI generated. SlopTracker, which attempts to track AI generated music on Spotify that is featured on official Spotify-curated playlists, that those artists are “draining” $0.1188 per second from real artists who could be making the music instead.

A judge caught a court reporter making AI-generated errors in a court transcript, and put court reporters everywhere on notice for their use of AI. 

In a memorandum decision concerning a case about a man who sold drugs to another man who overdosed and died, filed on July 23, Judge Paul Felix wrote in a footnote of the decision that a transcript contained errors that looked a lot like generative AI. The footnote was spotted by attorney Rob Freund on X.

Patreon laid off 93 employees, totaling 20 percent of its workforce on Thursday morning, according to an email to creators and staff from CEO Jack Conte.

In a message sent to everyone on the platform signed up as a creator, and posted to the site, with the subject line “A Painful Update about our Team,” Conte wrote that the business is “healthy and strong” and that the core business of Patreon is not changing. 

Conte included the email sent to Patreon employees in the message. In the email, he wrote that the company is undergoing both a “workforce reduction” and is “changing our organizational structure and how we work.” Specifically, he wrote, this means “we’re flattening the organization, refocusing teams on our top priorities, and evolving key aspects of our operations to make us faster at adapting to change.”

Conte is careful in this email to both express that he doesn’t view AI as a replacement for the human creativity the platform is built on and profits from — devoting a section of the email to saying as much — and also that AI is fundamentally “transforming” the tech industry, noting that it has an impact on how the company operates. 

“To be clear about the impact of AI on today’s decision: we are not making the above changes because we believe AI replaces humans,” Conte wrote. “The more we have learned to use these new tools, the clearer it has become that they are not substitutes for the creativity, judgment, detail orientation, or craftsmanship that our teammates have in spades, nor do they replace the desire for human connection that all of us cherish so deeply. That’s my personal opinion, but more importantly, it’s the foundation of Patreon’s strategy: our product vision and business are both predicated on the value of human creativity and human connection. AI has fundamentally transformed the tech industry, though, including how we work, how we build products, how we communicate, and more. That does have an impact on how we operate and organize.”

Earlier this month, Patreon announced that it’s partnering with Cloudflare to block crawlers from stealing creators’ work to train AI models. “I HAVE A KICKASS PRODUCT UPDATE FOR YOU ALL!” Conte wrote in a post on Instagram. “This is live and happening at the network level on all posts published on Patreon.” The company later elaborated on the partnership in a blog post.

Included in the internal email shared publicly are severance details for laid-off workers, including 16 weeks of pay, remaining on payroll through the company’s August 20 vesting date, one additional week of pay for every full year worked at the company, additional cash payments for recent hires who haven’t reached their one-year cliff, healthcare coverage through the end of the year for eligible employees and families, and a $1,500 stipend to replace their company laptops.

Long-range Ukrainian drones struck the Ilsky oil refinery in Russia's Krasnodar Krai on July 7, hitting one of the facility's technological workshops, a source in Ukraine's military intelligence (HUR) told the Kyiv Independent.

Located roughly 500 kilometers (311 miles) from Ukrainian-controlled territory, the refinery is among the largest in southern Russia, producing over 6 million tons of fuel annually.

It is involved in the reception, storage, and processing of hydrocarbons and distributes refined products via road and rail. The refinery is part of Russia's military-industrial complex and plays a direct role in supporting Moscow's war effort, the source said.

The Russian regional operational headquarters claimed that "drone debris" fell on the oil refinery.

The strike marks a renewed wave of Ukrainian attacks on Russian oil infrastructure, following a months-long pause since March. On July 1, Ukrainian drones struck the Saratovorgsintez oil refinery in Russia's Saratov Oblast.

Kyiv has targeted dozens of refineries, oil depots, and military-industrial sites since the start of Russia's full-scale invasion in 2022. Winter drone attacks forced at least four Russian refineries to temporarily shut down.

This is the second known strike on the Ilsky refinery. Ukrainian drones, operated by the Security Service (SBU) and Special Operations Forces (SSO), previously targeted the facility on Feb. 17, causing a fire.

Krasnodar Krai, a strategic region along Russia's Black Sea coast, has increasingly come under Ukrainian drone attacks as Kyiv extends the range of its strikes deep into Russian territory.

HUR publishes Russian military order, claims proof of Moscow increasing military footprint in Armenia

HUR first made the claim on July 5, saying Russia was increasing its forces at the Gyumri base to exert greater influence in the South Caucasus and “destabilize the global security situation.”

The Kyiv IndependentChris York