The CEO seemingly having an affair with the head of HR at his company at the Coldplay concert is a viral video for the ages, but it is also, unfortunately, emblematic of our current private surveillance and social media hellscape.

The video, which is now viral on every platform that we can possibly think of, has been covered by various news outlets, and is Pop Crave official, shows Andy Byron, the CEO of a company called Astronomer, with his arms around Astronomer’s head of HR, Kristen Cabot. The jumbotron cuts from one fan to this seemingly happy couple. They both simultaneously die inside; “Oh look at this happy couple,” Coldplay lead singer Chris Martin says. The woman covers her face and spins away. The man ducks out of frame. “Either they’re having an affair or they’re very shy,” Martin said. The camera pans to another company executive standing next to them, who is seemingly shaking out of discomfort.

It is hard to describe how viral this is at the moment, in a world in which so many awful things are occurring and in which nothing holds anyone’s attention for any length of time and in a world in which we are all living in our own siloed realities. “Andy Byron” is currently the most popular trending Google term in the United States, with more than double the searches of the next closest term. 

There are so many levels to this embarrassment—the Coldplay of it all, the HR violation occurring on jumbotron, etc—that one could likely write a doctoral dissertation on this 15 second video.

But the aftermath of this video demonstrates a phenomenon that we have now seen over and over again on all of social media, but on TikTok especially: The instant, gleeful doxing of people for reasons that run the spectrum from the very bad (being a Nazi) to the utterly benign (dancing weird at a Taylor Swift concert, being a hot security guard, standing in public). “TikTok help me find him / her” is now so common that it is a meme.

404 Media doesn’t know what’s going on in Byron or Cabot’s marriages and neither does anyone else posting about it; we have zero clue what’s going on behind the scenes or what is currently occurring in the personal lives of everyone involved.

It is worth briefly considering the dystopia of this situation and its aftermath. It is not clear exactly how the Astronomer CEO was initially identified, but we have seen numerous cases where TikTok commenters and creators use Pimeyes and other readily available, often free facial recognition and social media research tools to identify a person. 

The Astronomer CEO’s name, his wife’s name, the head of HR’s name, and the third company executive’s name and social media profiles are all over the TikTok comments and Reddit comments. His latest LinkedIn post was full of comments about the incident, left before he disabled comments and ultimately deleted the post. Commenters have pointed out that his wife has removed his last name from several of her social media profiles. 

Polymarket’s degenerate gamblers turned the scandal into a market. The site lets users bet on events with binary outcomes like whether or not Byron will lose his job as CEO of Astronomer. A gambler pays a few cents to vote “yes” or “no” on the outcome of the event and wins the difference up to a dollar on each vote if they guessed right. Each vote shifts the odds up and down.

Today Polymarket opened two new Byron-related bets. The first asks the question: “Andy Byron out as Astronomer CEO by next Friday?” As of this writing, the site is giving him a 40 percent chance of leaving. Polymarket doesn’t care if he resigns or gets fired, the bet pays out to its “yes” voters so long as he leaves.

The crueler bet, and the one with more activity, is the “Astronomer Divorce Parlay.” A parlay is a series or combination of bets, and this Polymarket listing requires both Byron and Cabot to get  divorced. An announcement will satisfy the bet, it doesn’t have to be an official filing. But it can come from Byron and Cabot themselves or their spouses. The image attached to the bet is a screenshot from the moment the Jumbotron filmed the couple embracing. 

As of this writing, the odds are 21 percent that both couples divorce.

Polymarket’s X account got in on the action. As both of the markets opened, the account posted about Byron and Cabot four times in a row. “Smart bet?” Polymarket asked above a picture showing social media photos of the pair listed next to the current odds that they’d divorce their spouses.

Brands are using Byron and Cabot’s face to build hype. The NEON film studio posted a picture of the couple on X to promote its upcoming horror movie Together. Chipotle commented on the story with a picture of its billboard that reads “It’s OK to Cheat.” Most chilling, the NYC Department of Sanitation used the viral moment to remind everyone that it has cameras everywhere.

Scrolling through a TikTok search for “CEO of Astronomer” (a TikTok-recommended search term) is full of wannabe influencers giving their own take on the saga, greenscreening the video behind their cut out faces with titles like “CHEATING COLDPLAY AFFAIR BOSTON CONCERT COUPLE UPDATE,” and “COLDPLAY EXPUSO A DOS AMANTES.” Fox News, Buzzfeed, and other huge accounts have posted the footage more or less unedited on their own accounts. Writing this article, we understand, puts us into this exact fray.

But one does not have to have sympathy or empathy for a CEO to see how this sort of thing could and often does go off the rails. This example is emblematic of the problem specifically because it’s easy to laugh at these people and because they’re doing something distasteful, but not illegal. The same technologies used to dox and research this CEO are routinely deployed against the partners of random people who have had messy breakups, attractive security guards, people who look “suspicious” and are caught on Ring cameras by people on Nextdoor, people who dance funny in public, and so on. There has been endless debate about the ethics of doxing cops and ICE agents and Nazis, and there are many times where it makes sense to research people doing harm on behalf of the state or who are doing violent, scary things in to innocent people. It is another to deploy these technologies against random people you saw on an airplane or who had a messy breakup with an influencer. And of course, these same technologies are regularly deployed by police and the feds against undocumented immigrants, regular people, and people wanting to visit the United States on tourist visas. 

“JUST IN: Astronomer CEO Andy Byron’s wife removes his last name from her Facebook profile,” a follow up post said next to a screenshot from Byron’s wife’s Facebook account. And in a reply: The odds of them both getting divorced this year are soaring. 52% chance.”

ICE officers are able to point their smartphone’s camera at a person and near instantaneously run their face against a bank of 200 million images, then pull up their name, date of birth, nationality, unique identifiers such as their “alien” number, and whether an immigration judge has determined they should be deported from the country, according to ICE material viewed by 404 Media.

The new material, which includes user manuals for ICE’s recently launched internal app called Mobile Fortify, provides granular insight into exactly how ICE’s new facial recognition app works, what data it can return on a subject, and where ICE is sourcing that data. The app represents an unprecedented linking of government databases into a single tool, including from the State Department, Customs and Border Protection (CBP), the FBI, and state records. It also includes the potential for ICE to later add commercially available databases that contain even more personal data on people inside the United States.

“This app shows that biometric technology has moved well beyond just confirming someone's identity. In the hands of ICE officers, it's becoming a way to retrieve vast amounts of data about a person on demand just by pointing a camera in their face,” Dave Maass, director of investigations at the Electronic Frontier Foundation (EFF), told 404 Media. “The more they streamline its use, the more they streamline its abuse. When an officer says, ‘papers please,’ you could choose to say nothing and face the consequences; with face recognition, your options are diminished.”

A memo sent to ICE officials describes Mobile Fortify as “Biometric Capture Made Easy.” Beyond facial recognition, it has a feature that allows ICE officers to use a smartphone camera to scan and save a person’s fingerprints with a feature it called “contactless fingerprints.” It says the app “allows officers to run remote identity checks in the field by capturing contactless fingerprints and facial images using their ICE-issued iPhones. The application runs facial images against over 200 million DHS and Department of State records.” 

A demonstration included in the material shows a person standing in front of a phone’s camera with an ellipse on the screen where their face will be scanned. “Searching Face,” the app says. 

The material adds that fingerprints are run against DHS’s central repository of biometrics. To capture these, the user manual tells ICE officers to “align the subject’s fingers inside of the blue box and hold until the camera has captured the prints.”

ICE officers can use the tool to perform what the material describes as a “Super Query.” This includes querying multiple datasets at once related to “individuals, vehicles, airplanes, vessels, addresses, phone numbers and firearms,” according to the memo.

According to screenshots of the app viewed by 404 Media, the Super Query feature can search the National Crime Information Center (NCIC), which is an FBI-maintained database containing “missing persons and criminal information;” the Nlets system which includes whether someone has an outstanding state warrant against them; CBP’s system that screens people at the border called TECS; and another CBP database called the Seized Assets and Case Tracking System (SEACATS) that tracks material previously seized by CBP and ICE.

After scanning a face, fingerprints, or entering details about an identity document, the app can return a “SCREENED HITS FOUND” message, according to the user manuals. Those results include the subject’s name, nationality, age and date of birth, and unique identifiers such as their “alien registration,” and “CIV ID,” which is a system issued by United States Citizenship and Immigration Services (USCIS). The results also include “Immig. Judge Decision,” presumably referring to whether an immigration judge has ruled on this person’s case, and “final order.” In the example given in the user manual, the immigration judge decision section says “remove.”

404 Media first revealed the existence of Mobile Fortify in June, based on leaked ICE emails, which showed ICE was repurposing data collected when people enter or exit the United States for enforcement inside the country.

Mobile Fortify differs from more well-known facial recognition tools, such as Clearview AI, because Clearview doesn’t necessarily have access to government databases. When a law enforcement officer uses Clearview, it checks a subject’s face against a massive archive of images Clearview has scraped from social media and the wider web. That can return someone’s name, potentially, and their social media presence. Mobile Fortify, meanwhile, takes someone’s identity and then burrows into DHS’s and other agencies' databases to surface much more information about the person, and information that only the government may possess.

Flight Manifests Reveal Dozens of Previously Unknown People on Three Deportation Flights to El Salvador

Hacked data obtained by 404 Media reveals dozens more people on deportation flights to El Salvador who are unaccounted for. “We have not heard from these people’s families, so I think perhaps even they don’t know,” one lawyer said.

404 MediaJason Koebler

The material indicates that Mobile Fortify may soon include data from commercial providers too. One section says that “currently, LexisNexis is not supported in the application.” LexisNexis is a massive data broker that collates data including peoples’ addresses, phone numbers, and associates. LexisNexis did not respond to a request for comment.

In videos posted to social media, ICE officers have repeatedly been seen pointing their phones in a way that suggested they were scanning or taking a photo of someone’s face. It is unclear whether the officers were using this tool or not.

Facial recognition technology can be inaccurate and lead to wrongful arrests. In one case, a Detroit man spent 30 hours in a jail cell because the technology thought he was a criminal, the New York Times previously reported.

“I worry ICE could be embracing a less accurate system because they’re indifferent to mistakes or even trying to build in systems that maximize chances of producing a ‘match’ that gives them a pro forma basis for stopping or detaining people, an especially accrue risk as ICE is being criticized in court for improper stops and racial profiling,” Jake Laperruque, deputy director of the Center of Democracy and Technology's Security and Surveillance Project, told 404 Media.

ICE has repeatedly made mistakes during its ongoing mass deportation effort, including detaining American citizens or raiding the wrong house. 

ICE was recently granted a new budget of $28 billion.

DHS did not respond to a request for comment. The FBI deferred a request for comment to ICE. The State Department did not respond.

The flight manifests for three legally contested deportation flights from Texas to El Salvador contain dozens of additional, unaccounted for passengers than a previously published Department of Homeland Security (DHS) list of people deported from the United States on those flights, 404 Media has learned. The additional people on the flight manifest have not been publicly acknowledged by the U.S. government in any way, and immigration experts who have been closely monitoring Trump’s deportation campaign say they have no idea where these people are or what happened to them. 404 Media is now publishing the names of these people. 

On March 15, the Trump administration deported more than 200 people on three aircraft to a megaprison in El Salvador. A judge blocked the deportations, but hours later the flights still landed in the country. It marked one of the major turning points of the administration’s mass deportation efforts, and signaled what was to come around the country—a lack of due process, authorities ignoring judge’s rulings, and deporting people on the flimsiest of pretenses. Soon after these flights, CBS News published an “internal government list” of people it said were deported to CECOT, the notorious El Salvadorian megaprison.

But in May, a hacker targeted GlobalX, the airline that operated these flights and shared the data with 404 Media. In addition to the names of people who were on the list CBS News published, the GlobalX flight manifests contain the names of dozens of people who were supposedly on the flights but whose status and existence has not been acknowledged by the U.S. government or previously reported in the press. 

“We have this list of people that the U.S. government has not formally acknowledged in any real way and we pretty much have no idea if they are in CECOT or someplace else, or whether they received due process,” Michelle Brané, executive director of Together and Free, a group that has been working with families of deported people, told 404 Media. “I think this further demonstrates the callousness and lack of due process involved and is further evidence that the US government is disappearing people. These people were detained and no one knows where they are, and we don't know the circumstances […] For almost all of these people, there’s no records whatsoever. No court records, nothing.” 

“[The government is] not disclosing it and they’ve presumably been sent to a prison or sent somewhere by the U.S. government on a plane and have never been heard from since,” she added. “We have not heard from these people’s families, so I think perhaps even they don’t know.”

Brané added that it remains entirely unclear whether all of these people were actually on the flights or why they were on the manifests. If they were indeed on the flights, it is unknown where they currently are. That uncertainty, and the unwillingness of the U.S. government to provide any clarity about these people, is a major problem, she said.

While the stories of some of the people deported on these flights have garnered a lot of attention, such as Kilmar Abrego Garcia, U.S. authorities have refused to reveal the names of everyone on board. 

While the whereabouts and circumstances of most of these people remain unknown, Brané’s organization used publicly available data to try to better understand who they are. In some cases, Together and Free was able to identify a few details about specific people on the manifest. For example, one person on the manifest appears to have been arrested by local police in Texas in late December on drug possession charges and is listed in arrest records as being an “illegal alien.” Another person was arrested in Nashville in February on charges of driving without a license. For many other people listed, there is no easily discernible public data about who they are or why they appeared on the flight manifest.

Several other people are on the flight manifests and do not appear on the CBS News list, but their identities had already become public because their families have filed lawsuits or have been looking for them on social media. These include Abrego Garcia and Ricardo Prada Vásquez, a man whose family said he was “disappeared” because he did not appear on any official, publicly published lists. After the New York Times published an article about his disappearance, the Trump administration said he was at CECOT, and 404 Media was able to find his name on the March 15 flight manifests. 

In Venezuela, the family of another man who appears on the flight manifests but not on the CBS News list, Keider Alexander Flores Navas, has been protesting his disappearance and demanding answers. In a TikTok video posted in March, his mother Ana Navas explains that they suddenly stopped hearing from Keider before the March 15 flights. She said she eventually heard he was in federal detention. Then, she saw a photo of him in CECOT amongst a group of other prisoners: “The thing that worried me the most was he was not on any list. But this photo is from El Salvador. Lots of family members here recognize their sons [in official CECOT photos]. That’s my son,” she says, the camera panning to a circled image of Keider in CECOT.

In another TikTok video posted in June, the mother of 21-year-old Brandon Sigaran-Cruz explains that he had been “disappeared for three months” with no news of his whereabouts. Sigaran-Cruz also appears on the flight manifest but not the CBS News list. 

 The U.S. government previously acknowledged that, along with more than 200 Venezuelan citizens, it deported 23 Salvadorans to El Salvador on the three March 15 flights. There is no formal list of the Salvadorans who were on the flight, and none of them appeared on the CBS News list, which included only Venezuelan citizens. 

The United Nations’ Human Rights Office has also filed court petitions saying that it is investigating the “involuntary disappearances” of at least four Venezuelans who were sent to El Salvador on these flights. “Neither the Government of El Salvador nor the Government of the United States has published official information on the list of deported persons or their current place of detention,” the United Nations said in a “Report on Enforced or Involuntary Disappearances” it filed in court.

“There continues to be very little clarity as to the fate and whereabouts of the Venezuelans removed to El Salvador. To date, no official lists of the deported detainees have been published. Provision of further information by authorities is key, including providing families and their counsel with available information on the specific situation and whereabouts of their loved ones,” Elizabeth Throssell, a spokesperson for the UN Human Rights Office, told 404 Media in an email. “The UN Human Rights Office has been in contact with family members of over 100 Venezuelans believed to have been deported to El Salvador.”

404 Media asked the Department of Homeland Security (DHS) over multiple weeks if the agency had any legitimate security concerns with these names being published, or if it could tell us anything about these people. The agency never responded, despite responding to requests for comment for other 404 Media articles. GlobalX did not respond to a request for comment either.

“It is critical that we know who was on these March 15 flights,” Lee Gelernt, a lawyer at the American Civil Liberties Union (ACLU) and the lead counsel on the ACLU’s related case, told 404 Media. “These individuals were sent to a gulag-type prison without any due process, possibly for the remainder of their lives, yet the government has provided no meaningful information about them, much less the evidence against them. Transparency at a time like this is essential.”

In recent months, the U.S. government has said that the El Salvadorian government has jurisdiction over the people detained in CECOT, while El Salvador told the United Nations that “the jurisdiction and legal responsibility for these persons lie exclusively with the competent foreign authorities [the United States],” leading to a situation where people are detained in a foreign prison but both responsible parties are not willing to claim legal responsibility for them. A similar situation has happened in Florida at the “Alligator Alcatraz” camp, where people detained by the federal government are being held in a state-run facility, and experts have said it’s not clear who is in charge. Brané said with the massive increase in ICE funding as part of Trump’s new law, we are likely to see more detention camps, more detainments, more deportation flights, and, likely, more people who aren’t publicly accounted for in any way.

“When you look at what ICE is doing now in terms of how they treat people, how they operate when they're given even a little bit of rope, it’s terrifying to think what the budget increase is going to do,” Brané said. “This is a taste of what we're going to see on a much larger scale."

You can read the list below. 404 Media has removed people listed on the flight manifests as “guards” (404 Media found at least one of these names matched someone who lists their employment online as a flight transport detention officer). Reportedly eight women deported to El Salvador were later returned. 404 Media is not publishing the names of women known to have returned to the U.S. The manifest also includes the names of several El Salvadorians mentioned as being deported in a White House Press release, court proceedings, and media reports. We have not included their names below because the administration has formally acknowledged that they were deported.

Manuel Quijada-Leon
Irvin Quintanilla-Garcia
Jose Ramirez-Iraheta
Josue Rivera-Portillo
Jorge Rodriguez Gomez
Mario Jeavanni Rojas
Edgar Leonel Sanchez Rosales
Brandon Sigaran-Cruz
Miguel Enriquez Saravia
Abraham Hernandez-Mania
Jean Morales-Loaiza
Nelson Alfaro-Orellana
Jhonnarty Pachecho-Chirinos
Cristian Alpe-Tepas
Jordyn Alexander Alvarez
Jose Alvarez Gonzalez
Wilfredo Avendano Carrizalez
Jose Gregorio Buenano Cantillo
Istmar Campos Mejia
Jose Chanta-Ochoa
Keider Alexander Flores Navas
Noe Florez-Valladares
Miguel Fuentes-Lopez
Roberto Interiano Uceda
Jose Lopez Cruz
Diego Maldonado-Fuentes
William Martinez-Ruano
Osmer Mejias-Ruiz
Iran Ochoa Suescun
David Orantez Gonzalez
Ariadny Araque-Cerrada
Elena Cuenca Palma
Maria Franco Pina
Mayerkis Guariman Gonzalez
Wilmary Linares-Marcano
Scarlet Mendoza Perez
Ofreilimar Peña Boraure
Edilianny Stephany Rivero Sierralta
Dioneli Sanz Aljorna
Anyeli Sequera Ramirez
Yanny Suarez Rodriguez
Karla Villasmil-Castellano

A major mystery about a long-lost legend that was all the rage in Medieval England but survives in only one known fragment has been solved, according to a study published on Tuesday in The Review of English Studies.

Roughly 800 years ago, a legend known today as the Song of Wade was a blockbuster hit for English audiences. Mentions of the heroic character showed up in the works of Geoffrey Chaucer, for example. But the tale vanished from the literature centuries later, puzzling generations of scholars who have tried to track down its origin and intent. 

Now, for the first time, researchers say they’ve deciphered its true meaning—which flies in the face of the existing interpretation. 

“It is one of these really interesting and very unusual situations where we have a legend that was widely known and hugely popular throughout the Middle Ages, and then very suddenly in the middle of the 16th century, in the High Renaissance, it's just completely lost,” said James Wade, fellow in English at Girton College, University of Cambridge, who co-authored the study.

In 1896, the Medieval scholar M.R. James made a breakthrough on this literary cold case when he discovered a fragment of the Song of Wade in the Humiliamini sermon, which is part of a compendium that dates back to the 12th century. James brought the text to his colleague Israel Gollancz, a philologist with expertise in early English literature, and together they worked on a translation.

It is “the only surviving fragment” of the Song of Wade, said co-author Seb Falk, fellow in history and philosophy of science also at Girton College. “Obviously it had been around already for a while by the time this was written because it's part of the culture. The person who writes this sermon clearly expects his listeners to understand it and to know what he's talking about.”

James and Gollancz “knew there was no surviving text and they understood what they were looking at,” added Wade. “This was big news. It made the papers in 1896.”

But while the 19th century scholars recovered the sermon, their translation only deepened the mystery of the enigmatic text. For instance, references to “elves” and “sprites” in the translation suggest that the Song of Wade falls into a genre of fantastical epics about supernatural monsters. But when Chaucer references Wade in his works Troilus and Criseyde and The Merchant’s Tale, he places the character in a totally different tradition of chivalric romances which are rich with metaphors, but typically favor more grounded scenarios.

Chaucer’s mentions of Wade have perplexed scholars for centuries: in 1598, for instance, an early Chaucer editor named Thomas Speght wrote: “Concerning Wade and his [boat] called Guingelot, as also his strange exploits in the same, because the matter is long and fabulous, I passe it over.” In other words, Speght didn’t even try to decipher what Chaucer meant with his references to Wade.

This punt has become legendary in literary circles. “F. N. Robinson wrote in 1933 that Speght’s comment ‘has often been called the most exasperating note ever written on Chaucer’, and Richard Firth Green observed that Speght’s note ‘has caused generations of scholars to tear their hair out,’” Wade and Falk write in their new study.  

In 1936, the scholar Jack Bennett "supposed that there is ‘probably no better known crux in Chaucer than the tale of Wade,’” the pair added.

A few years ago, Wade and Falk set out to see if they could shed light on this famous and persistent riddle. Like so many good ideas, it began over a lunchtime conversation. From there, the team slowly and methodically worked through the sermon, scrutinizing each letter and rune.  

“Just trying to decipher the thing took quite a lot of work—transcribing it and then making an initial translation,” Falk said. “But I started realizing there's some really interesting material here from my point of view, as a historian of science, with lots of animals mentioned, both in Middle English and in Latin.”

As the pair worked through the text, they began to suspect that the scribe who originally copied the work may not have been very familiar with Middle English, leading to some transcription errors with certain runes. In particular, they found that the longstanding translation of “elves” and “sprites” were, in their view, more likely to be “wolves” and “sea snakes.” This transforms a key Song of Wade passage, “Some are elves and some are adders; some are sprites that dwell by waters,” to: 

“Some are wolves and some are adders; some are sea-snakes that dwell by the water.”

It may seem like a subtle shift, but it is a major sea change for the interpretation of the text. The switch to animals, as opposed to supernatural beings, suggests that the preacher who wrote the sermon was using animals as metaphors for human vices and behaviors—a reading that provides a much better fit for the overall sermon, and at last explains why Chaucer viewed Wade in the tradition of chivalry.

“It became pretty clear to us that the scribe had probably made some kind of a mistake because he was used to writing Latin rather than English,” Falk said. “We're not here to say that people who read it differently previously were stupid to miss it, but I think by looking at it from the outside in, we got a different perspective.”

“It then radically changes the meaning of the passage from being about monsters to being about animals, and therefore, changes it from being a piece about mythical beasts to a piece about courtly romance,” he added.

In addition to relieving centuries-old headaches over the legend, Wade and Falk also speculate that the sermon was originally written by English poet and abbot Alexander Neckam (1157–1217), based on its style and context clues. 

For the researchers, the thrill of the discovery lies not only in decoding the fragment, but in recovering a missing piece of cultural memory. While their new translation and attribution to Neckam are both tentative and may be disputed by other scholars, the study still opens a window into a long-lost legend and shows how fresh eyes can uncover insights in even the most perplexing fragments.

“By putting a completely different slant on it and, we think, understanding it properly, we have come much closer to the true meaning of the Wade legend,” Falk said. “Now, obviously we've only got three lines of this presumably much longer poem, and therefore, we can't pretend that we understand the thing in full, but I think we can understand it much better than we ever have before.”

“It reminds us that as time moves on, there's always the possibility of generational amnesia, of forgetting things, of losing things, and that when you have a chance to get a little bit back of something that humanity has lost, or that culture has lost, it's a really exciting moment,” Wade concluded.

Steam, the dominant digital storefront for PC games operated by Valve, updated its guidelines to forbid “certain kinds of adult content” and blamed restrictions from payment processors and financial institutions. The update was initially spotted by SteamDB.info , a platform that tracks and publishes data about Steam, and reported by the Japanese gaming site Gamespark.

The update is yet another signal that payment processors are lately becoming more vigilant about what online platforms that host adult content they’ll provide services to and another clear sign that they are currently the ultimate arbiter of what kind of content can be made easily available online, or not. 

Steam’s policy change appears under the onboarding portion of its Steamworks documentation for developers and publishers. The 15th item on a list of “what you shouldn’t publish on Steam” now reads: “Content that may violate the rules and standards set forth by Steam’s payment processors and related card networks and banks, or internet network providers. In particular, certain kinds of adult only content.”

It’s not clear when exactly Valve updated this list, but an archive of this page from April shows that it only had 14 items then. Other items that were already on the list included “nude or sexually explicit images of real people” and “adult content that isn’t appropriately labeled and age-gated,” but Valve did not previously mention payment processors specifically. 

"We were recently notified that certain games on Steam may violate the rules and standards set forth by our payment processors and their related card networks and banks," Valve spokesperson Kaci Aitchison Boyle told me in an email. "As a result, we are retiring those games from being sold on the Steam Store, because loss of payment methods would prevent customers from being able to purchase other titles and game content on Steam. We are directly notifying developers of these games, and issuing app credits should they have another game they’d like to distribute on Steam in the future."

Valve did not respond to questions about where developers might find more details about payment processors’ rules and standards. 

SteamDB.info, which also tracks when games are added or removed from Steam, noted many adult games have been removed from Steam in the last 24 hours. Sex games, many of which are of very low quality and sometimes include very extreme content, have been common on Steam for years. In April, I wrote about a “rape and incest” game called No Mercy which the developers eventually voluntarily removed from Steam after pressure from users, media, and lawmakers in the UK. The majority of games I saw that were removed from Steam recently revolve around similar themes, but we don’t know if they were removed by the developers or Valve, and if they were removed by Valve because of the recent policy change. Games are removed from Steam every day for a variety of reasons, including expired licensing deals or developers no longer wanting to support a game. 

However, Steam’s policy change comes at a time that we’ve seen increased pressure from payment processors around adult content. We recently reported that payment processors have forced two major AI models sharing platforms, Civitai and Tensor.Art, to remove certain adult content.

Update: This story has been updated with comment from Valve. 

We start this week with a series of articles from Emanuel about a crackdown in the AI industry. After the break, Sam tells us about the ‘Save Our Signs’ campaign which hopes to preserve the history of national parks. In the subscribers-only section, Jason rants about how AI will not save the media industry. 

Listen to the weekly podcast on Apple Podcasts, Spotify, or YouTube. Become a paid subscriber for access to this episode's bonus content and to power our journalism. If you become a paid subscriber, check your inbox for an email from our podcast host Transistor for a link to the subscribers-only version! You can also add that subscribers feed to your podcast app of choice and never miss an episode that way. The email should also contain the subscribers-only unlisted YouTube link for the extended video version too. It will also be in the show notes in your podcast player.

• 404 Media Los Angeles event details (free for subscribers; $10 otherwise)
• a16z-Backed AI Site Civitai Is Mostly Porn, Despite Claiming Otherwise
• Hugging Face Is Hosting 5,000 Nonconsensual AI Models of Real People
• Payment Processors Are Pushing AI Porn Off Its Biggest Platforms
• 'Save Our Signs' Wants to Save the Real History of National Parks Before Trump Erases It
• The Media's Pivot to AI Is Not Real and Not Going to Work

Yesterday I ordered my lunch from an AI operating a drive-thru. It was fine. Banal. Boring even. A new experience that I think will become routine in the future.

The AI drive-thru operator isn’t cutting edge tech deployed in an upscale market to win over high value consumers. I live at the edge of a South Carolina city with a little more than 140,000 people. A booming metropolis with the best and the finest, it is not.

There’s a lot of local fast food fried chicken joints here and one of them is Bojangles. It’s mid. Better than KFC and not as good as Popeyes, Bojangles is fine if you’re hungry but you’ll forget the meal as soon as it’s done and you’ll never yearn for it. Last year the restaurant said it would deploy an AI agent at its drive-thru windows. It’s called, I shit you not, Bo-Linda and made by the Israeli tech firm Hi-Auto.

According to the Bojangles website, “Bo-Linda™ can take guest orders 96+% of the time with no human intervention,” and “improve overall satisfaction by offloading order taking from team members and providing a consistent guest experience.”

When Bo-Linda finally arrived in South Carolina, I went to see what the fuss was about. It was crushingly dull. A preview of a time in the near future, I think, when the AI bubble retracts and the agents are common. It took my order with an efficiency that, I’ll be honest, is not typical of the typical fast food worker. The worst part was its constant attempts to up-sell me.

“Do you want to upgrade your drink to our new water-melon iced tea?” It asked.

“No thank you.”

“Would you like to add our new peach cobbler for $1.99?”

“No thank you.”

“May I get you anything else?”

“No, that’s it.”“Would you like to round up for military scholarships?”“No thank you.”

“You’re welcome. Thank you. Your total is $10.89.”

When 404 Media founder Joseph Cox watched the video of my interactions, he made fun of my “no thank yous.” What can I say? There’s an ingrained and often stifling politeness that’s bred into us in the American South. Even though I knew I was talking to a machine, I couldn’t not be nice to it.

My thought in the immediate aftermath is that the whole thing was painless. My order wasn’t complicated, but it was correct. The machine never stumbled over itself or asked for clarification. It knew what I wanted and the humans at the window gave it to me. A few conversations with friends and a quick scan of social media in the area show that other people have had much the same interactions with Bo-Linda.

The drive-thru AI, much like the chicken it sold me, is fine. Forgettable.

It was later, sitting at home, and doing a little research for the story that concerns popped up. OpenAI CEO Sam Altman has said that saying “please” and “thank you” to ChatGPT has cost the company tens of millions of dollars. How much water and energy had I burned being polite to Bo-Linda the chatbot?

Sometimes it feels like the answers to these questions don’t matter. We’re barreling forward into the AI future, whether we like it or not. Data centers are springing up across America and nuclear power plants are coming back online, so Bojangles can make a little more money and so people in the drive-thru can feel a little less friction before eating their meal.

This is how a new technology takes over, what it feels like right before it becomes ubiquitous. One day you wake up and the cameras are everywhere, able to recognize your face and chart your movements across the city you live in. One day you look up and everyone has their face buried in their phone. It happened by degrees, but so gradually you didn’t notice. There were signs along the way, dangers and warnings.

But mostly, it was fine, as boring and routine as ordering chicken at a drive-thru.

So-called 3D-printed ghost guns are untraceable firearms that can be assembled at home. But cutting edge work from a forensic expert in California and researchers at the University of Oklahoma may soon show investigators can trace a 3D printed object to the specific printer that made it.

Weapons manufactured using 3D printers have been a subject of Biden-era legislation and recent Supreme Court scrutiny. It’s possible to download the blueprints for a firearm and build it in your home. There’s no serial number to track and no store to scrutinize your purchase. Luigi Mangione used a ghost gun to allegedly assassinate United Healthcare CEO Brian Thompson.

For a while, I have said that the AI slop endgame, for social media companies, is creating a hyper personalized feed full of highly specific content about anything one could possibly imagine. Because AI slop is so easy to make and because social media algorithms are so personalized, this means that Facebook, Instagram, TikTok, or YouTube can feed you anything they perceive its users to possibly want. So this means that AI slop makers are exploring ever more niche areas of content. 

Case in point: Facebook AI slop about the horrific and deadly Texas flood. Topical AI content about disasters, war, current events, and news stories are at this point so commonplace that they are now sadly barely notable, and AI-powered “misinformation” about horrible events are all over every social media feed I can think of. But as we document our descent into this hellhole, I thought some AI slop surfaced on Bluesky by Christina Stephens was particularly notable:

Yesterday, Silicon Valley trade publication The Information launched TITV, a live-streaming news program. For its first episode, the company’s founder Jessica Lessin interviewed Mark Zuckerberg about his recent attempts to hire AI executives away from other companies for huge sums of money.

It is impossible to say what actually happened in the interview, however, because there were audio problems with the livestream—the Zuckerberg segment was completely silent, giving us a minute and 50 seconds of video of Zuckerberg’s face moving with no sound coming out.

In recent weeks, however, Zuckerberg has really ramped up his rhetoric about “superintelligence,” throwing hundreds of millions of dollars at AI researchers in attempts to hire them at Meta, and beginning work on massive data centers around the country to power his AI. Zuckerberg and Meta have talked endlessly about AI as a game changer, while being unable or unwilling to keep AI-generated misinformation and slop off of Meta platforms.

Meanwhile, there has been increased discussion about the possible power of using AI for journalism, which we have written about extensively. I saw the silent interview with Zuckerberg as both an opportunity to check out the awesome abilities of Meta’s AI, and to test AI tools at a task some of them are supposedly good at—reading lips. If these technologies worked as well as their creators have promised, it might be possible to provide the world with information from The Lost Mark Zuckerberg Interview.

Here is how my experiment went.

Because Zuckerberg was obviously talking about Meta AI, I thought I would try his tool first. I took video from the interview and asked Meta AI to lip read it to tell me what they were saying:

“I’d love to help! However, I’m a text-based AI and don’t have capability to visually lip read videos,” Meta AI said. It then suggested that I “find a professional lip reader or transcription service that specializes in lip reading videos” or “use automated lip reading software.” 

I then tried to upload still images to see if Meta AI could parse them:

When I uploaded additional images, Meta AI was also not able to provide any information about what was being said.

I then went to ChatGPT, because Zuckerberg is reportedly offering pay packages of up to $300 million to OpenAI staffers to come work at Meta. I uploaded the 1:50 video and ChatGPT told me “the video processing took too long and timed out.” I then uploaded a 25 second clip and it told me “the system is still timing out while trying to extract frames.” I then asked it to do the first five seconds and it said “even with the shorter clip and smaller scope (first 5 seconds), the system timed out.” I then asked for it to extract one single frame, and it said “it looks like the system is currently unable to extract even a single frame from the video file.” ChatGPT then asked me to take a screenshot of Zuckerberg. I sent it this: 

And ChatGPT said “the person appears to be producing a sound like ‘f’ or ‘v’ (as in ‘video’ or ‘very’),” but that “possibly ‘m’ or ‘b,’ depending on the next motion.” I then shared the 10 frames around that single screenshot, and ChatGPT said “after closely analyzing the progression of lip shapes and facial motion,” the “probable lip-read phrase” was “This is version.” I then uploaded 10 more frames and it said the “full phrase so far (high confidence): ‘This version is just.’”

I then decided to try to extract every frame from the video and upload it to ChatGPT.

I went to a website called frame-extractor.com and cut the video into 3,000 frames. After it had processed 700 of them, I tried to upload them to ChatGPT and it did not work. I then decided I would go 10 frames at a time from the beginning of the clip. Even though I sent an entirely different portion of the video and told ChatGPT we were starting from a different part of the video, it still said that the beginning of the video said “this version is.” I continued uploading frames, 10 at a time. These frames included both Lessin and Zuckerberg, not just Zuckerberg.

ChatGPT slowly began to create a surely accurate transcript of the lost audio of this interview: “This version is just that it we built,” ChatGPT said. As I added more and more frames, it refined the answer: “This version is what we’re going to do,” it said. Finally, it seemed to make a breakthrough. “Is this version of LLaMA more powerful than the one we released last year?” the ChatGPT transcript said. It was not clear about who was speaking, however. ChatGPT said "her mouth movements," but then explained that the "speaker is the man on the left" (Lessin, not Zuckerberg, was speaking in these frames). 

I had uploaded 40 of a total of 3,000 frames. Zoom video is usually 30 fps, so in approximately 1.5 seconds, Lessin and/or Zuckerberg apparently said “Is this version of LLaMA more powerful than the one we released last year?” I then recorded this phrase at a normal speaking speed, and it took about four seconds. Just a data point.

Lipreadtest

0:00

/4.973333

1×

I then got an error message from ChatGPT, and got rate-limited because I was uploading too much data. It told me that I needed to wait three hours to try again. 

Finally, I did what Meta AI told me to do, and tried a bespoke AI lip reading app. I found one called ReadTheirLips.com, which is powered by Symphonic Labs. This is a tool that people have been trying to use in recent months to figure out what Donald Trump and Jeffrey Epstein were saying to each other in silent b-roll news footage, without much success.

I paid $10 for three minutes worth of transcription and asked it to lip read using its “Multiface Detection.” After waiting 10 minutes, I got an error message that said “Transcription failed, no credits have been used, try again later.” I then asked it to focus only on Zuckerberg, and actually got some text. I separately asked it to focus on Lessin.

Here is a transcript of what the AI says they were talking about. It has not been edited for clarity and I have no idea which parts, if any, are accurate:

LESSIN: Thanks for joining us again, TV. We're happy to have you already this morning. News that you've spent even more money with your big announcement about your new supercomputers. We'll get to that, but to start, you've been in huge scale like I.

ZUCKERBERG: Happy TO BE HERE. We're GOING TO TALK A LITTLE BIT ABOUT META'S AI STRATEGY. It's BEEN BUSY, YOU KNOW? I THINK THE MOST EXCITING THING THIS YEAR IS THAT WE'RE STARTING TO SEE EARLY GLIMPSES OF SELF-IMPROVEMENT WITH THE MODELS, WHICH MEANS THAT DEVELOPING SUPERINTELLIGENCE IS NOW.

LESSIN: You HAVE BEEN ON A PLANE OF AI HIRING, WHY AND WHY NOW? 

ZUCKERBERG: Insight, and we just want to make sure that we really strengthen the effort as much as possible to go for it. Our mission with a lab is to deliver personal superintelligence to everyone in the world, so that way, you know, we can put that power in every individual's hand. I'm really excited about it.

LESSIN: I DON'T KNOW, I DON'T KNOW, I DON'T KNOW.

ZUCKERBERG: Than ONE OF THE OTHER LABS YOU'RE DOING, AND YOU KNOW MY VIEW IS THAT THIS IS GOING TO BE SOMETHING THAT IS THE MOST IMPORTANT TECHNOLOGY IN OUR LIVES. IT'S GOING TO UNDERPIN HOW WE DEVELOP EVERYTHING AND THE COMPANY, AND IT'S GOING TO AFFECT SOCIETY VERY WISELY. SO WE JUST WANT TO MAKE SURE WE GET THE BEST FOCUS.

LESSIN: Did YOU FEEL LIKE YOU WERE BEHIND WHAT WAS COMING OUT OF LAW BEFORE I'M NOT ADJUSTING.

ZUCKERBERG: On THIS FROM ENTREPRENEURS TO RESEARCHERS TO ENGINEERS WORKING ON THIS HIDDEN INFRASTRUCTURE, AND THEN OF COURSE WE WANT TO BACK IT UP WITH JUST AN ABSOLUTELY MASSIVE AMOUNT OF COMPUTER RESEARCH, WHICH WE CAN SUPPORT BECAUSE WE HAVE A VERY STRONG BUSINESS MODEL THAT THROWS OFF A LOT OF CAPITAL. LET'S TALK ABOUT.

LESSIN: Like THIS SUMMER, PARTICULARLY, YOU SWITCH GEARS A LITTLE BIT.

ZUCKERBERG: I THINK THE FIELD IS ACCELERATING, YOU KNOW, WE KEEP ON TRACK FOR WHERE WE WANT TO BE, AND THE FIELD KEEPS US MOVING FORWARD.

The video ends there, and it cuts back to the studio.

Update: The Information provided 404 Media with several clips (with audio) from Lessin's interview with Zuckerberg, as well as a real transcript of the interview. Here is the real segment of what was said. As you can see, the AI captured the jist of this portion of the interview, and actually did not do too bad:

Lessin: Mark, thanks for joining TITV. We're happy to have you here. Already this morning, [there’s] news that you've spent even more money with your big announcement about your new supercomputers. We'll get to that. But to start, you took a huge stake in ScaleAI. You have been on a blitz of AI hiring. Why, and why now?

Zuckerberg: Yeah, it's been busy. You know, I think the most exciting thing this year is that we're starting to see early glimpses of self-improvement with the models, which means that developing super intelligence is now in sight, and we just want to make sure that we really strengthen the effort as much as possible to go for it. Our mission with the lab is to deliver personal super intelligence to everyone in the world, so that way we can put that power in every individual's hand. And I'm really excited about it. It's a different thing than what the other labs are doing.

And my view is that this is going to be something that is the most important technology in our lives. It's going to underpin how we develop everything at the company, and it's going to affect society very widely. So we just want to make sure that we get the best folks to work on this, from entrepreneurs to researchers to engineers working on the data and infrastructure.

And then, of course, we want to back up with just an absolutely massive amount of compute which we can support, because we have a very strong business model that throws off a lot of capital.

Lessin: Did you feel like you were behind coming out of Llama 4? It seems like this summer, in particular, you switched gears a little bit.

Zuckerberg: I think the field is accelerating, you know, we keep on having goals for where we want to be. And then the field keeps on moving faster than we expect.

The rest of the interview is available at The Information.

ICE Block, an app that lets users warn others about the location of ICE officers, and which for a short while was the top of the social media App Store chart, does protect users’ privacy and doesn’t share your location with third parties, according to a recent analysis from a security researcher. ICE Block already claimed that it did not collect any data from the app; the analysis now corroborates that.

“It’s not uploading your location at all, when you make a report that report isn’t associated with your device in any way, and there are no third party services that it talks to or sends data to,” Cooper Quintin, senior public interest technologist at the Electronic Frontier Foundation (EFF), who analyzed the ICE Block app, told 404 Media.

Hugging Face, a company with a multi-billion dollar valuation and one of the most commonly used platforms for sharing AI tools and resources, is hosting over 5,000 AI image generation models that are designed to recreate the likeness of real people. These models were all previously hosted on Civitai, an AI model sharing platform 404 Media reporting has shown was used for creating nonconsensual pornography, until Civitai banned them due to pressure from payment processors. 

Users downloaded the models from Civitai and reuploaded them to Hugging Face as part of a concerted community effort to archive the models after Civitai announced in May it will ban them. In that announcement, Civitai said it will give the people who originally uploaded them “a short period of time” before they were removed. Civitai users began organizing an archiving effort on Discord earlier in May after Civitai indicated it had to make content policy changes due to pressure from payment processors, and the effort kicked into high gear when Civitai announced the new “real people” model policy. 

At the time of writing, the Discord channel has hundreds of members who are still finding and sharing models that have been removed from Civitai and are reuploading them to Hugging Face. Some users have even shared a piece of software, also hosted on Hugging Face, which allows users to automatically upload Civitai models to Hugging Face in batches. 

Hugging Face did not respond to multiple requests for comment. It also did not respond to specific questions about how and if it plans to moderate these models given the fact that they were previously hosted on a platform primarily used for AI generating pornography, and which our reporting shows were used to create noncensual pornography. 

I found the Civitai models of real people that were reuploaded to Hugging Face thanks to a paper I covered where researchers scraped Civitai. The paper showed that the platform was primarily used for pornographic content, and that it deleted at least 50,000 AI models designed to recreate the likeness of real people once it changed its policy in May. The researchers, Laura Wagner and Eva Cetinic from the University of Zurich, provided me with a spreadsheet of all the deleted models, which included the name of the models (which is almost always the name of a female celebrity or lesser known internet personality), a link to where it was previously hosted on Civitai, and the SHA256 hash Civitai uses to identify all the models hosted on its site. 

The people who are reuploading the Civitai models to Hugging Face are seemingly trying to hide the purpose of those models on Hugging Face. On Hugging Face, these models have generic names and URLs like “LORA” or “Test model.” Users can’t tell that these models are used to generate the likeness of real people just by looking at their Hugging Face page, nor would they be able to find them by searching for the names of celebrities on Hugging Face. In order to find them, users can go to a separate website the Civitai archivists created. There, they can enter the name of a Civitai model, the link where it used to be hosted on Civitai before it was deleted, or the model’s SHA256 hash. All of these will lead users to a page which explains what the model is, show its name, as well as several images showing the kind of images it can generate. At the bottom of that page is a link to one or more Hugging Face “mirrors” where the model has been reuploaded. 

By using Wagner’s and Cetinic’s data and entering it into this Civitai archive site, I was able to find the Civitai models hosted on Hugging Face. 

Hugging Face’s content policy bans “Unlawful, defamatory, fraudulent, or intentionally deceptive Content (e.g., disinformation, phishing, scams, inauthentic behavior),” as well as “Sexual Content used for harassment, bullying, or created without explicit consent.” Models that generate the likeness of real people don’t have to be used for unlawful or defamatory ends, and they only produce sexual content if people choose to use them that way. There’s nothing in Hugging Face’s content policy that explicitly forbids AI models that recreate the likeness of real people. 

However, the Hugging Face Ethics & Society group, which is “committed to operationalizing ethics at the cutting-edge of machine learning,” has identified six “high-level categories for describing ethical aspects of machine learning work,” one of which is that AI should be “Consentful.”

“Consentful technology supports the self-determination of people who use and are affected by these technologies,” the company explains. Examples of this, the company says, includes “Avoiding extractive, chauvinist, ‘dark,’ and otherwise ‘unethical’ patterns of engagement.”

Other AI models that recreate the likeness of real people could conceivably not violate any of these principles. For example, two of the deleted Civitai models that were reuploaded to Hugging Face were designed to recreate the likeness of Vladimir Putin, which in theory people would want to use in order to mock or criticize the Russian president. However, the vast majority of the models are of female celebrities, which my reporting has shown is being used to create nonconsensual sexual content, and which were deleted en masse from Civitai because of pressure from payment processors who didn’t want to be associated with that type of media. 

In the two years that I’ve been reporting about Civitai, a platform for sharing AI image generation models that has been instrumental in the production of AI generated non-consensual porn, Civitai has consistently argued that the amount of adult content on the site has been overstated. But new research shows that, if anything, the amount of adult content on Civitai has been underestimated.

In their paper, “Perpetuating Misogyny with Generative AI: How Model Personalization Normalizes Gendered Harm,” researchers Laura Wagner and Eva Cetinic from the University of Zurich studied more than 40 million user-generated images on Civitai and over 230,000 models. They found “a disproportionate rise in not-safe-for-work (NSFW) content and a significant number of models intended to mimic real individuals” on the platform, they write in the paper.

“What began as a promising creative breakthrough in TTI [text-to-image] generation and model personalization, has devolved into a pipeline for the large-scale production of sensational, biased, and abusive content. The open-source nature of TTI technologies, proclaimed as a democratizing force in generative AI, has also enabled the propagation of models that perpetuate hypersexualized imagery and nonconsensual deepfakes,” Wagner and Cetinic write in their paper. “Several indicators suggest a descent into a self-reinforcing feedback loop of platform decay. These include a dramatic increase in NSFW imagery, from 41% to 80% in two years, as well as the community’s normalization of deepfakes, misogynistic tropes, and other exploitative content.”

To visualize just how dominant adult content was on Civitai, check the chart below, which shows the distribution of images by “NSFW browsing levels” over time. These categories, which are inspired by the Motion Picture Association film rating system and are used by Civitai to tag images, show that adult content was always a significant portion of all images hosted on the site, but that the portion of “overtly sexual, or disturbing” content only grew as the site became more popular, and exploded starting in 2024. The chart is based on Civitai’s own numbers and categorization system which the researchers scraped from the site. It likely undercounts the number of explicit images on the site since as both the researchers and I observed during my reporting, not all adult content is tagged as such. 

In December, 2023, Civitai CEO Justin Maier told Venture Beat that “less than 20% of the posted content is what we would consider ‘PG-13’ or above.” When I reached Maier for comment for this article, he told me that “The VentureBeat figure cited a December 2023 snapshot, when adult posts were a minority. The mix shifted in 2024 as many NSFW creators migrated from platforms that no longer allow that content.”

However, the data in the paper shows that by October of 2023, 56 percent of all images on the site were tagged as “NSFW” and were designated by Civitai as “PG-13” or above.

In May, Civitai announced it’s banning all AI image generation models designed to recreate the likeness of real people because of pressure from payment processors. Since the authors of the paper were already tracking hundreds of thousands of models hosted on Civitai, they could easily see which models were removed, giving us a first clear look at how common those models were. 

Overall, they saw that more than 50,000 models designed to AI-generate the likeness of real people were removed because of the ban. These are models that Civitai itself tagged as “person of interest,” the tag it uses to indicate a model recreates the likeness of a real person, so the actual number of models depicting real people is likely higher. 

It’s hard to say if the most popular AI models on Civitai were all popular just because they were used to generate explicit images, because people could use models tagged as NSFW to generate non-nude images and vice versa. For example, according to the data collected by the researchers the most popular AI image generation model on Civitai was EasyNegative with almost 600,000 downloads. It’s not tagged or promoted as a model for generating pornography, but images that users created with it, which are shared on its Civitai model page, show it is commonly used that way. 

Other very popular models on Civitai are clearly designed to generate explicit images. The sixth most popular model with 360,000 downloads is Nudify XL: Better Bodies, which its creator says is for “nude female frontals.” A model called Realistic Vaginas - God Pussy 1 had 256,000 downloads. The POV Squatting Cowgirl LoRA model, which Civitai tagged as a “sex” model, had 189,000 downloads. 

The authors of the paper also conducted deeper analysis of the 40,000 most downloaded models on Civitai. In the 11,151 models where they could extract textual training data, meaning text that indicates what kind of images the models were trained on, they found “specifically abusive terms.” 5.6 percent included the keywords “loli” (558 models) and/or “shota” (69 models), Japanese terms commonly used to refer to sexualized depictions of pre-pubescent girls and boys. About 2.1 percent (189 models) included the keyword “rape.”

The data shows with clear numbers what we have long argued at 404 Media: adult content drives technological innovation and early adoption, and this has been especially true in the world of generative AI. Despite its protestation to the contrary, Civitai, which is one of the fastest growing platforms in that industry, and that the influential Silicon Valley venture capital firm Andreessen Horowitz invested in, grew because of explicit content, much of which was nonconsensual. 

“The rapid rise of NSFW content, the over-representation of young female subjects, and the prioritization of sensational content to drive engagement reflect an exploitative, even abusive dynamic,” the researchers wrote. “Additionally, structural discrimination embedded in today’s open-source TTI tools and models have the potential to cause significant downstream harm as they might become widely adopted and even integrated into future consumer applications.” 

Adult content driving innovation and early adoption doesn’t have to be harmful. As the researchers write, it’s the choices platforms like Civitai make that give us these outcomes. 

“The contingent nature of technology, shaped by online communities, platform operators, lawmakers, and society as a whole, also creates opportunities for intervention,” they write. “Model-sharing hubs and social media platforms both have the capacity to implement safeguards that can limit the spread of abusive practices such as deepfake creation and abusive imagery.”

Many trains in the U.S. are vulnerable to a hack that can remotely lock a train’s brakes, according to the U.S. Cybersecurity and Infrastructure Security Agency (CISA) and the researcher who discovered the vulnerability. The railroad industry has known about the vulnerability for more than a decade but only recently began to fix it.

Independent researcher Neil Smith first discovered the vulnerability, which can be exploited over radio frequencies, in 2012. 

“All of the knowledge to generate the exploit already exists on the internet. AI could even build it for you,” Smith told 404 Media. “The physical aspect really only means that you could not exploit this over the internet from another country, you would need to be some physical distance from the train [so] that your signal is still received.”

The Moderate Party of Sweden has removed an AI tool from its website after people used it to generate videos of Prime Minister Ulf Kristersson asking Adolf Hitler for support.The tool allowed users to generate videos of Kristersson holding an AI-generated message in an attempt to promote the candidate ahead of the general election in Sweden next year.

Swedish television station TV4 used the tool to generate a video of Kristersson on a newspaper above the headline “Sweden needs Adolf Hitler” after it noticed that it had no guardrails or filters.

In the video TV4 generated using the website, Kristersson makes his pitch over stock footage of old people embracing. A woman runs through a field, the camera focusing on flowers while the sun twinkles in the background. Cut to Kristersson. He turns a blue board around. “We need you, Adolf Hitler,” it says.

The Moderates removed the AI system from its website, but the videos of Ulf asking Hitler to join the Moderates remain on social media and TV4’s website..

In an attempt to bolster its party's ranks, Moderates launched a website that allowed users to generate a custom video of Kristersson asking someone to join the party. The idea was probably to have party members plug in the names of friends and family members and share what appeared to be a personalized message from the PM asking for their support.

In the video, Kristersson stands in front of stairs, makes his pitch, and turns around a blue tablet that bears a personalized message to the viewer. The system apparently had no guardrails or filters and Swedish television station TV4 was able to plug in the names Adolf Hitler, Ugandan dictator Idi Amin, and Norwegian mass murderer Anders Breivik.

The Moderate Party did not return 404 Media’s request for a comment about the situation, but told TV4 it shut down the site as soon as it learned people were using it to generate messages with inappropriate names.

The Moderate Party’s AI-generated video was simple.. It filmed the PM holding a blue board it could easily overlay with input from a user and then used AI to generate the fake newspaper and a few other slides. Preventing people from typing in “Hitler” or “Anders Brevik” would have been as simple as maintaining a list of prohibited names, words, and phrases, something that every video game and service does. Users are good at bypassing guardrails, but the Moderate’s AI tool appeared to have none.

Users making content you don’t want to be associated with is one of the oldest and most well known problems in AI. If you release a chatbot, generative photo system, or automated political greeting generator, someone will use it to reference the Nazis or make nonconsensual porn.

When Microsoft launched TAY in 2016, users turned it into a Hitler-loving white nationalist in a few hours. Eight years later, another Microsoft AI product had a loophole that let people make AI-generated nudes of Taylor Swift. Earlier this year, Instagram’s AI chatbots lied about being licensed therapists.

As immigration raids roll out across the U.S., those affected are processing the experience in the normal 2025 way—via vertical video. 

Across social media, people are uploading clips with uncanny-valley titles like “A normal day for me after being deported to Mexico” and “3 things I wish I knew before self-deporting from the US!” These posts have the normal shape, voiceovers, and fonts of influencer content, but their dystopian topic reflects the whiplash of the current historical moment. 

Doomscrolling last week, a particular clip caught my eye. A man sits on the bottom bunk of a metal bed, staring down at the floor, with the caption “Empezando una nueva vida después de que me Deportaran a México” (“Starting a new life after being Deported to Mexico”).

On May 23, we got a very interesting email from Ghost, the service we use to make 404 Media. “Paid subscription started,” the email said, which is the subject line of all of the automated emails we get when someone subscribes to 404 Media. The interesting thing about this email was that the new subscriber had been referred to 404 Media directly from chatgpt.com, meaning the person clicked a link to 404 Media from within a ChatGPT window. It is the first and only time that ChatGPT has ever sent us a paid subscriber.

From what I can tell, ChatGPT.com has sent us 1,600 pageviews since we founded 404 Media nearly two years ago. To give you a sense of where this slots in, this is slightly fewer than the Czech news aggregator novinky.cz, the Hungarian news portal Telex.hu, the Polish news aggregator Wykop.pl, and barely more than the Russian news aggregator Dzen.ru, the paywall jumping website removepaywall.com, and a computer graphics job board called 80.lv. In that same time, Google has sent roughly 3 million visitors, or 187,400 percent more than ChatGPT. 

This is really neither here nor there because we have tried to set our website up to block ChatGPT from scraping us, though it is clear this is not always working. But even for sites that don’t block ChatGPT, new research from the internet infrastructure company CloudFlare suggests that OpenAI is crawling 1,500 individual webpages for every one visitor that it is sending to a website. Google traffic has begun to dry up as both Google’s own AI snippets and AI-powered SEO spam have obliterated the business models of many media websites. 

This general dynamic—plummeting traffic because of AI snippets, ChatGPT, AI slop, Twitter no workie so good no more—has been called the “traffic apocalypse” and has all but killed some smaller websites and has been blamed by executives for hundreds of layoffs at larger ones. 

Despite the fact that generative AI has been a destructive force against their businesses, their industry, and the truth more broadly, media executives still see AI as a business opportunity and a shiny object that they can tell investors and their staffs that they are very bullish on. They have to say this, I guess, because everything else they have tried hasn’t worked, and pretending that they are forward thinking or have any clue what they are doing will perhaps allow a specific type of media executive to squeeze out a few more months of salary.

But pivoting to AI is not a business strategy. Telling journalists they must use AI is not a business strategy. Partnering with AI companies is a business move, but becoming reliant on revenue from tech giants who are creating a machine that duplicates the work you’ve already created is not a smart or sustainable business move, and therefore it is not a smart business strategy. It is true that AI is changing the internet and is threatening journalists and media outlets. But the only AI-related business strategy that makes any sense whatsoever is one where media companies and journalists go to great pains to show their audiences that they are human beings, and that the work they are doing is worth supporting because it is human work that is vital to their audiences. This is something GQ’s editorial director Will Welch recently told New York magazine: “The good news for any digital publisher is that the new game we all have to play is also a sustainable one: You have to build a direct relationship with your core readers,” he said.

Becoming an “AI-first” media company has become a buzzword that execs can point at to explain that their businesses can use AI to become more ‘efficient’ and thus have a chance to become more profitable. Often, but not always, this message comes from executives who are laying off large swaths of their human staff.

In May, Business Insider laid off 21 percent of its workforce. In her layoff letter, Business Insider’s CEO Barbara Peng said “there’s a huge opportunity for companies who harness AI first.” She told the remaining employees there that they are “fully embracing AI,” “we are going all-in on AI,” and said “over 70 percent of Business Insider employees are already using Enterprise ChatGPT regularly (our goal is 100%), and we’re building prompt libraries and sharing everyday use cases that help us work faster, smarter, and better.” She added they are “exploring how AI can boost operations across shared services, helping us scale and operate more efficiently.” 

Last year, Hearst Newspapers executives, who operate 78 newspapers nationwide, told the company in an all-hands meeting audio obtained by 404 Media that they are “leaning into [AI] as Hearst overall, the entire corporation.” Examples given in the meeting included using AI for slide decks, a “quiz generation tool” for readers, translations, a tool called Dispatch, which is an email summarization tool, and a tool called “Assembly,” which is “basically a public meeting monitor, transcriber, summarizer, all in one. What it does is it goes into publicly posted meeting videos online, transcribes them automatically, [and] automatically alerts journalists through Slack about what’s going on and links to the transcript.”

Welcome back to the Abstract! Here are the studies that caught my eye this week.

First up, a bummer! NASA is facing devastating cuts to Earth science, and science in general, which is pretty important for an agency tasked with understanding the universe. I try to keep this newsletter relatively downer-free, but current events are not cooperating with this aim.  

Then: the early bird gets the comet, a donkey destined for decapitation, a grand universal theory of coolness, and the case of the stolen exomoons.  

Knowing about the planet we live on is good, actually

Millet, Dylan et al. “NASA Earth Science Division provides key data.” Science.

This may sound obvious for the esteemed readers of this newsletter, but it apparently bears repeating: Earth, our home planet and the only known life-bearing world in the vast expanse of spacetime, is worthy of some passing interest.

This is true because—and I can’t stress this enough—we live on planet Earth. All our bones and guts and snacks are here, so it’s probably wise to get the lay of the land. But as a bonus, Earth is wildly interesting, a restless substrate unlike anything we have seen in our own solar system or beyond it. 

Despite these considerations, the Trump administration plans to gut NASA’s Earth Science Division (ESD), the world leader in world-watching. In a letter published in Science, researchers from the recently dissolved NASA Earth Science Advisory Committee lamented the administration’s proposed budget cut of more than 50 percent to ESD, warning that it “would come at a profound cost to US society and scientific leadership.”

“NASA ESD accounted for just 0.03 percent of US spending in 2024,” said researchers led by Dylan Millet of the University of Minnesota. “This investment returns its value many times over by improving predictions, by spurring technological innovation and high-tech jobs, and by forging the knowledge of the planet that is needed for short- and long-term planning.”

“The budget cuts proposed for ESD would cancel crucial satellites that observe Earth and its atmosphere, gut US science and engineering expertise, and potentially lead to the closure of NASA research centers,” the team said. “Given that the cuts would prevent the US from training and preparing the next generation of the scientific and technical workforce, the consequences would be long-lasting.”

This is just the latest appeal from scientists on behalf of NASA, which is also facing catastrophic cuts to its overall Science Mission Directorate (SMD), the arm that oversees ESD. Last week, every past administrator of the SMD, the agency's top job for science leadership, signed a letter urging Congress to reject the cuts of about 47 percent to the directorate. 

“Each one of us knows what it’s like to shepherd an ambitious project forward, knowing that its payoff will come years after we have left the agency,” the administrators said. “This proposed budget ends nearly all future investments for both new missions and advanced technology for science. It walks away from dozens of current, extraordinarily successful and productive science missions in extended operations on a combined budget that is only about three percent of NASA’s annual funding.”

Fortunately, the US Senate appropriations committee has voted in favor of a bill rejecting the science cuts, but it has a long road to go down before taking effect, with plenty of opportunity to fall apart. 

Needless to say, turning a blind eye to Earth at a time when our activities are reshaping its climate and biosphere, would be a huge loss. As one last twist of the knife, Trump just gave the top job at NASA to the guy from The Real World—all while ignoring the actual real world. 

In other news…

Mark your calendars for July 2061

Barbieri, Cesare et al. “Preparing for the 2061 return of Halley’s comet. A rendezvous mission with an innovative imaging system.” Planetary and Space Science.

Prepare for the return of everyone’s favorite space iceball: Comet Halley. Scientists have flagged the comet’s next visit as arriving in the summer of 2061 and proposed an audacious space rendezvous with Halley on its wild ride toward the Sun. 

“Although the crucial phases of the comet’s ingress in the inner Solar System are still more than 30 years in the future, we started to examine the feasibility of a space mission using present-day rockets and technologies,” said researchers led by Cesare Barbieri of the University of Padova.

God bless the astro-preppers. Sure, this event will occur decades into the future, in the twilight of the millennials. But they make a pretty good case that we should get moving if we want to take full advantage of the iconic visitor, which “will be better positioned for observation from terrestrial observers than during the 1985–1986 apparition, as it will be on the same side of the Sun as the Earth.”

“We stress that a concerted effort is needed in the current decade to plan and approve a rendezvous mission to [Comet Halley],” the team concluded. “Indeed, the scenario here described needs launches before 2040, less than 15 years from now.”

The last days of a sacrificial ass 

Arnold, Elizabeth et al. “An isotopic perspective on equid selection in cult at Tell eṣ-Ṣâfi/Gath, Israel.” PLOS ONE.

Who would have guessed that a study about a Bronze Age donkey corpse would be a tear-jerker? Researchers have shed new light on female donkeys (or jennies), which were imported from Egypt to the Tell es-Safi/Gath site in Israel some 5,000 years ago for ritual purposes. 

One specimen, called EQ1, is particularly noteworthy because it was decapitated and had its limbs tied together, unlike all the other donkeys buried at the site. "It is evident the animal was sacrificed, the head entirely cut off and carefully placed on the abdomen facing in the opposite direction,” said researchers led by Elizabeth Arnold of Grand Valley State University.  

“It can be deduced that even though EQ1 was grazed locally toward the end of her life, she was treated slightly differently from the other local equids,” the team continued. “This imported donkey was kept penned and foddered with hay that was harvested in the valley, a product of dry farmed cereals. This donkey was never herded with other livestock east of the site.”

The unique treatment of EQ1 suggests that the “Egyptian donkey might have been seen as an exotic and special animal, worthy of specific ritual use,” the study concluded. While it’s truly impressive that so much about this jenny can be inferred from her bones, there’s also an eerie pathos to imagining the animal hanging out for months, receiving preferential treatment, unaware of the sand flowing through the hourglass.    

It’s a real cool club, and you’re not part of it

Pezzuti, Todd and Warren Caleb et al. “Cool People,” Journal of Experimental Psychology.

Science has invested its prodigious powers into the ultimate social mystery: What makes a person “cool”? Is it putting “cool” in scare quotes? (No!). Researchers have now developed a working theory of coolness by asking nearly 6,000 people in Australia, Chile, China, Hong Kong, Germany, India, Mexico, Nigeria, Spain, South Africa, South Korea, Turkey, and the United States to define this much-coveted attribute.

The results revealed six main traits associated with cool people, which were distinct from traits linked with “good” people. “Cool people are perceived to be more extraverted, hedonistic, powerful, adventurous, open, and autonomous, whereas good people are more conforming, traditional, secure, warm, agreeable, universalistic, conscientious, and calm,” said authors Todd Pezzuti of Universidad Adolfo Ibañez, Caleb Warren of the University of Arizona, and Jinjie Chen of the University of Georgia.

“This pattern is stable across countries, which suggests that the meaning of cool has crystallized on a similar set of values and traits around the globe,” the team said.  

There you go, the cheat code to coolness. I’m exhausted just reading it.

Grand Theft Moons

Dencs, Zoltán et al. “Grand theft moons: Formation of habitable moons around giant planets.” Astronomy & Astrophysics.

The winner of Best Study Title this week goes to “Grand theft moons,” which explores how stars might steal moons from their own planets, and whether these “exomoons” could be habitable. The study models the formation of exomoons around giant gas worlds at various distances from their stars, measured in astronomical units (au), where one au is the distance between Earth and the Sun. 

Planets with orbits of one or two au are more likely to sport exomoons in the habitable zone, but they are also at risk of stars yanking the exomoons away in brazen acts of “stellar theft,” according to the study.

“Our simulations show that moons with masses between Mars and Earth could form around planets with masses about ten times that of Jupiter, and many of these moons could potentially be habitable at 1 − 2 au stellar distances,” said researchers led by Zoltán Dencs of the Gothard Astrophysical Observatory, “These findings suggest that it is worth investigating not only rocky planets but also gas giants for Earth-like habitable environments.”

In addition to raising some fascinating questions, let’s hope this study inspires Rockstar to take the GTA franchise to outer space. I want to throw an astronaut off a lunar buggy and take it for a joyride across the Moon.

Thanks for reading! See you next week.

Tensor.Art, an AI image creating and model sharing site announced on Friday that it is “temporarily” restricting AI models, tools, and posts related to pornographic content or the depiction of real-world celebrities due to pressure from payment processors.

The announcement is yet another example of payment processors acting as the ultimate arbiter of what kind of content can be easily made available online and those companies’ seemingly increased focus on AI-generated adult or nonconsensual content. 

The news is especially significant following a similar change in policy from Civitai, an AI model sharing platform 404 Media reporting has shown was used for creating nonconsensual pornography. After Civitai banned AI models designed to generate the likeness of real people and certain types of adult content in May, many Civitai users and model creators migrated their models to Tensor.Art. The announcement listed three items in the “Scope of Impact” of the decision: Banning “NSFW” content, banning content based on real-world celebrities, and  temporarily disabling its “Civitai Import” feature, which allowed people to easily move their Civitai models to Tensor.Art.

“We fully understand that this is very frustrating for many creators and users 😞,” Tensor.Art said in its announcement on Discord. “Due to the previous controversy over real-person content on Civitai, TA [Tensor.Art] has unfortunately been affected because of the ‘Civitai import’ feature. Owing to mandatory requirements from credit card organizations and regulatory authorities, we are compelled to make this temporary decision.”

Tensor.Art also listed the “Reasons for Adjustment” as:

Tensor.Art said that these changes will take place within the next 72 hours, and asked model creators to clarify if their models are “safe for work” in order to “prevent unintended impact.”

It’s not clear what Tensor.Art will look like or what its policies will be at the end of this “temporary” period. Civitai made similar changes permanently and still hasn’t been able to renew service from its payment processing providers or find new ones. Tensor.Art, however, is suggesting it’s not ready to give up on that type of content. 

“This is not the end,” Tensor.Art said in the announcement. “We are actively seeking solutions to minimize the impact of these restrictions and exploring compliant ways to restore currently hidden content. We remain committed to our original mission.”

Tensor.Art did not immediately respond to a request for comment.  

This is Behind the Blog, where we share our behind-the-scenes thoughts about how a few of our top stories of the week came together. This week, we discuss choosing what to cover, fishing expeditions, and the library.

JOSEPH: There’s an interesting cybersecurity story going on: a politically-motivated, right wing extremist hacker broke into Columbia University, stole swathes of applicant and other data, and then leaked parts of it. That’s how you got this New York Times article about New York mayoral hopeful Zohran Mamdani applying to the university as Asian and African American. Predictably, some people are reading that in bad faith, when in reality it shows how stupid and regimented application processes can be in a diverse society.

This week The Verge published a piece called The Columbia hack is a much bigger deal than Mamdani’s college application, which basically elaborates on that point. One section was this:

And yet, there has been precious little reporting on the Columbia hack. Wired hasn’t covered it, and, until this story, neither has The Verge. Nor have The Chronicle of Higher Education, CyberScoop, 404 Media, TechCrunch, or Krebs on Security. These—including The Verge—are small to medium-size entities, and there’s any number of possible reasons why they didn’t pick it up. (On our end, it was partly because we were short-staffed during a national holiday, and partly because we didn’t immediately piece together how extraordinary this particular hack is.) But coverage at the much bigger, well-resourced institutions is also scanty. The Wall Street Journal passed on the story. Reuters has a brief on the initial outage; AP has a short write-up as well, which The Washington Post ran as part of their syndication deal.