The National Institutes of Health claims it’s being strained by an onslaught of AI-generated research applications and is capping the number of proposals researchers can submit in a year.

In a new policy announcement on July 17, titled “Supporting Fairness and Originality in NIH Research Applications,” the NIH wrote that it has recently “observed instances of Principal Investigators submitting large numbers of applications, some of which may have been generated with AI tools,” and that this influx of submissions “may unfairly strain NIH’s application review process.” 

“The percentage of applications from Principal Investigators submitting an average of more than six applications per year is relatively low; however, there is evidence that the use of AI tools has enabled Principal Investigators to submit more than 40 distinct applications in a single application submission round,” the NIH policy announcement says. “NIH will not consider applications that are either substantially developed by AI, or contain sections substantially developed by AI, to be original ideas of applicants. If the detection of AI is identified post award, NIH may refer the matter to the Office of Research Integrity to determine whether there is research misconduct while simultaneously taking enforcement actions including but not limited to disallowing costs, withholding future awards, wholly or in part suspending the grant, and possible termination.” 

Starting on September 25, NIH will only accept six “new, renewal, resubmission, or revision applications” from individual principal investigators or program directors in a calendar year. 

Earlier this year, 404 Media investigated AI used in published scientific papers by searching for the phrase “as of my last knowledge update” on Google Scholar, and found more than 100 results—indicating that at least some of the papers relied on ChatGPT, which updates its knowledge base periodically. And in February, a journal published a paper with several clearly AI-generated images, including one of a rat with a giant penis. In 2023, Nature reported that academic journals retracted 10,000 "sham papers," and the Wiley-owned Hindawi journals retracted over 8,000 fraudulent paper-mill articles. Wiley discontinued the 19 journals overseen by Hindawi. AI-generated submissions affect non-research publications, too: The science fiction and fantasy magazine Clarkesworld stopped accepting new submissions in 2023 because editors were overwhelmed by AI-generated stories.

According to an analysis published in the Journal of the American Medical Association, from February 28 to April 8, the Trump administration terminated $1.81 billion in NIH grants, in subjects including aging, cancer, child health, diabetes, mental health and neurological disorders, NBC reported.

Just before the submission limit announcement, on July 14, Nature reported that the NIH would “soon disinvite dozens of scientists who were about to take positions on advisory councils that make final decisions on grant applications for the agency,” and that staff members “have been instructed to nominate replacements who are aligned with the priorities of the administration of US President Donald Trump—and have been warned that political appointees might still override their suggestions and hand-pick alternative reviewers.” 

NIH told 404 Media: "NIH developed this policy to ensure that the research application system promotes fairness and originality and to mitigate the potential overload of its review systems. NIH has observed cases of large numbers of applications (e.g., more than 40 being submitted in a single application submission round) which are likely evidence of use of AI tools in application development. As AI is becoming increasingly commonplace, it is reasonable to anticipate a potentially rapid increase in the use of AI in application development. A thorough analysis of application trends was conducted to determine the number of applications per calendar year the NIH should accept from a principal investigator. Specifically, the analysis revealed that 1.3% of PIs submitted more than six applications in 2024. Based on our data, this limit will affect a relatively small number of applicants."

Updated 4:00 p.m. EST to include comment from NIH.

Conservative content mill PragerU is partnering with the White House to make AI-generated videos of founding fathers and Revolutionary War-era randos.

PragerU is a nonprofit organization with a mission “to promote American values through the creative use of digital media, technology and edu-tainment,” according to its website. It’s been criticized for advancing climate denial and slavery apologism, frequently publishes videos critical of “wokeness” and “DEI,” and is very concerned about “the death of the West.” It has also been increasingly integrated into school curricula around the country.

PragerU held a launch event for the series, “Road to Liberty,” on June 25. Secretary Linda McMahon took some time away from dismantling the Department of Education to speak at the event. In person at the White House, visitors can tour a display of notable Revolutionary War people and places, and scan a QR code on displays that take them to PragerU’s AI-generated videos of people from that time period speaking. 

Each of the videos highlights a different person who was alive during the signing of the Declaration of Independence, from former presidents to relatively minor players in the fight for independence. The videos are clearly AI-generated, with the sepia-toned peoples’ mouths moving almost independently from the rest of their faces in some of them. In one, an AI-generated John Adams says “facts do not care about our feelings,” a phrase commonly attributed to conservative commentator and PragerU contributor Ben Shapiro. 

At the end of the videos, there's a logo for the White House with the text "brought to you by PragerU," and a disclaimer: "The White House is grateful for the partnership with PragerU and the U.S. Department of Education in the production of this museum. This partnership does not constitute or imply U.S. Government or U.S. Department of Education endorsement of PragerU."

Professor of history Seth Cotlar spotted the videos in a thread on Bluesky: 

I asked Cotlar, as someone who specializes in American history and the rise of the far-right, what stood out to him about these videos. I thought it was odd, I said, that they chose to include people like politician and disgraced minister Lyman Hall and obscure poet Francis Hopkinson alongside more well-known figures like John Adams or Thomas Jefferson. 

“You're right to note that it's a pretty odd collection of figures they've chosen,” Cotlar said. “My guess is that this is part of the broader right wing populist push to frame themselves as the grassroots ‘true Americans,’ and they're including all of these lesser known figures with the hopes that their viewers will be like ‘oh wow, look at all of these revolutionary freedom fighters like me who were just kinda ordinary guys like me but who still changed history.’” 

He also said it’s noteworthy that the “Road to Liberty” lineup so far is almost entirely white men, including the random dudes like Hall and Hopkinson. “The lack of any pretense to inclusion is pretty notable. Even conservative glosses on the Revolution from the pre-Trump era would have included things like the Rhode Island Regiment or Lemuel Haynes or Phyllis Wheatley. Needless to say, they absolutely do not include Deborah Sampson,” Cotlar said. All of the people in the “coming soon” section on PragerU’s website are also white men. 

AI slop has become the aesthetic of the right, with authoritarians around the world embracing ugly, lazy, mass-produced content like PragerU’s founding father puppets. Here in the U.S., we have President Donald Trump hawking it on his social media accounts, including AI-generated images of himself as the Pope and “Trump Gaza,” an AI video and song depicting the West Bank as a vacation paradise where Trump parties alongside his former bestie Elon Musk. As Republicans used the response to Hurricane Helene to blame migrants, Amy Kremer, founder of Women for Trump, posted an AI image of a child caught in a flood hugging a puppy and then said she didn’t care that it wasn’t real: “Y’all, I don’t know where this photo came from and honestly, it doesn’t matter,” she wrote on X. Mike Lee shared the same image. AI slop makes for quick and easy engagement farming, and now it’s being produced in direct partnership with the White House.

I’m not sure what app or program PragerU is using to make these videos. I thought, at first, that they might be using one of the many basic lipsyncing or “make this old photo come alive” mobile apps on the market now. But the videos look better, or at least more heavily produced, than most of those apps are capable of. Just to make sure they haven’t somehow advanced wildly in the last few months since I checked one out, I tried one of them, Revive, and uploaded an image of John Adams to see if it would return anything close to what PragerU’s putting out. It did not. 

The PragerU videos aren't this bad, but they also aren’t as good as what would come out of Veo 3, the newest AI video generator, which generates highly realistic videos complete with sound and speech, from text prompts. I gave Veo a painting of John Adams and told it what to say; PragerU probably isn’t using this generator, because the result is much more realistic than what’s in the “Road to Liberty” series, even when I use a screenshot from one of their videos.

On the off chance the culprit is Midjourney—although the series’ style and the way the subjects’ mouths move almost independently of the rest of their faces don’t match what I’ve seen of Midjourney’s videos—I tried that one, too. I just gave Midjourney the same Adams portrait and a prompt for it to animate him praising the United States and it returned a raving lunatic, silently screaming. 

Striking out so far, I emailed Hany Farid, a professor at UC Berkeley and Chief Science Officer of synthetic media detection company GetReal, and asked if he had any leads. He said it looked similar to what comes out of AI video creation platform HeyGen, which creates AI talking heads and generates speech for them using ElevenLabs. I tried this on screenshots of the avatars in PragerU’s Martha Washington and John Adams videos to see if the puppet-mouth-style matched up, and they were pretty close.

PragerU’s videos are still more heavily produced than what I could make using the free version of HeyGen; it’s possible they used a combination of these to make the videos, plus some old-fashioned video editing and animation to create the final products. PragerU reported almost $70 million in income last year, they can afford the effort. 

“While the PragerU stuff is distinctly terrible, it's not like our culture has commemorated the Revolution with high-minded sophistication,” Cotlar told me. “I was 8 during the bicentennial and while I definitely learned some stuff about the founding era, most of what I absorbed was pretty schlocky.” He mentioned the "Bicentennial minutes" that were broadcast in 1975 and 76, sponsored by Shell, and which TV critic John J. O’Connor called “so insubstantial as to be almost meaningless.” The series won an Emmy.

In the last two years, several states, beginning with Florida, have approved PragerU content to be taught in public school classrooms. In Oklahoma, teachers relocating from states with "progressive education policies” will have to undergo an assessment in partnership with PragerU to determine if they’re allowed to teach. "If you want to teach here, you'd better know the Constitution, respect what makes America great, and understand basic biology,” State Superintendent Ryan Walters said in a press release. “We're raising a generation of patriots, not activists, and I'll fight tooth and nail to keep leftist propaganda out of our classrooms."

Data preservationists and archivists have been working tirelessly since the election of President Donald Trump to save websites, data, and public information that’s being removed by the administration for promoting or even mentioning diversity. The administration is now targeting National Parks signs that educate visitors about anything other than “beauty” and “grandeur,” and demanding they remove signs that mention “negative” aspects of American history. 

In March, Trump issued an executive order, titled “Restoring Truth and Sanity To American History,” demanding public officials ensure that public monuments and markers under the Department of the Interior’s jurisdiction never address anything negative about American history, past or present. Instead, Trump wrote, they should only ever acknowledge how pretty the landscape looks.

Last month, National Park Service directors across the country were informed that they must post surveys at informational sites that encourage visitors to report "any signs or other information that are negative about either past or living Americans or that fail to emphasize the beauty, grandeur, and abundance of landscapes and other natural features," as dictated in a May follow-up order from Interior Secretary Doug Burgum. QR codes started popping up on placards in national parks that take visitors to a survey that asks them to snitch on "any signs or other information that are negative about either past or living Americans or that fail to emphasize the beauty, grandeur, and abundance of landscapes and other natural features." 

On one side of the world, a very online guy edits a photo of then-Vice President Nominee JD Vance with comically-huge and perfectly round chipmunk cheeks: a butterfly flaps its wings. A year later, elsewhere on the planet, a Norwegian tourist returns home, rejected from entry to the U.S. because—he claims—border patrol agents found that image on his phone and considered the round Vance meme “extremist propaganda.”

“My initial reaction was ‘dear god,’” the creator of the original iteration of the meme, Dave McNamee, told me in an email, “because I think it's very bad and stupid that anyone could purportedly be stopped by ICE or any other government security agency because they have a meme on their phone. I know for a fact that JD has these memes on his phone.”

On Monday, Norwegian news outlets reported that Mads Mikkelsen, a 21-year-old tourist from Norway, claimed he was denied entry to the United States when he arrived at Newark International Airport because Customs and Border Patrol agents found "narcotic paraphernalia" and "extremist propaganda" on his phone. Mikkelsen told Nordlys that the images in question were a photo of himself with a homemade wooden pipe, and the babyface Vance meme. (The meme he shows on his phone is a version where Vance is bald, from the vice presidential debate.)

McNamee posted his original edit of Vance as a round-faced freak in October 2024. "For every 100 likes I will turn JD Vance into a progressively apple cheeked baby,” he wrote in the original X post. In the following months, Vance became vice president, the meme morphed into a thousand different versions of the original, and this week is at the center of an immigration scandal.

It’s still unclear whether Mikkelsen was actually forbidden entry because of the meme. Mikkelsen, who told local outlets he’d been detained and threatened by border agents, showed the documentation he received at the airport to Snopes. The document, signed by a CBP officer, says Mikkelsen “is not in possession of a valid, un-expired immigrant visa,” and “cannot overcome the presumption of being an intending immigrant at this time because it appears you are attempting to engage in unauthorized employment without authorization and proper documentation.” 

The U.S. Department of Homeland Security (DHS) wrote in social media posts (and confirmed to 404 Media), "Claims that Mads Mikkelsen was denied entry because of a JD Vance meme are FALSE. Mikkelsen was refused entry into the U.S. for his admitted drug use." Hilariously, DHS and Assistant Secretary Tricia McLaughlin reposted the Vance meme on their social media accounts to make the point that it was NOT babyface Vance to blame.

Earlier this week, the State Department announced that visa applicants to the U.S. are now required to make their social media profiles public so the government can search them. 

“We use all available information in our visa screening and vetting to identify visa applicants who are inadmissible to the United States, including those who pose a threat to U.S. national security. Under new guidance, we will conduct a comprehensive and thorough vetting, including online presence, of all student and exchange visitor applicants in the F, M, and J nonimmigrant classifications,” the State Department said in an announcement. “To facilitate this vetting, all applicants for F, M, and J nonimmigrant visas will be instructed to adjust the privacy settings on all of their social media profiles to ‘public.’”

The meme is now everywhere—arguably more widespread than it ever was, even at its peak virality. Irish Labour leader Ivana Bacik held it up during an address concerning the U.S.’s new visa rules for social media. Every major news outlet is covering the issue, and slapping Babyface Vance on TV and on their websites. It’s jumped a news cycle shark: Even if the Meme Tourist rumor is overblown, it reflects a serious anxiety people around the world feel about the state of immigration and tourism in the U.S. Earlier this month, an Australian man who was detained upon arrival at Los Angeles airport and deported back to Melbourne claimed that U.S. border officials “clearly targeted for politically motivated reasons” and told the Guardian agents spent more than 30 minutes questioning him about his views on Israel and Palestine and his “thoughts on Hamas.”  

Seeing the Vance edit everywhere again, a year after it first exploded on social media, has to be kind of weird if you’re the person who made the Fat Cheek Baby Vance meme, right? I contacted McNamee over email to find out. 

When did you first see the news about the guy who was stopped (allegedly) because of the meme? Did you see it on Twitter, did someone text it to you...

MCNAMEE: I first saw it when I got a barrage of DMs sending me the news story. It's very funny that any news that happens with an edit of him comes back to me. 

What was your initial reaction to that?

MCNAMEE: My initial reaction was "dear god," because I think it's very bad and stupid that anyone could purportedly be stopped by ICE or any other government security agency because they have a meme on their phone. I know for a fact that JD has these memes on his phone.  

What do you think it says about the US government, society, ICE, what-have-you, that this story went so viral? A ton of people believed (and honestly, it might still be the case, despite what the cops say) that he was barred because of a meme. What does that mean to you in the bigger picture?

MCNAMEE: Well I think that people want to believe it's true, that it was about the meme. I think it says that we are in a scary world where it is hard to tell if this is true or not. Like 10 years ago this wouldn’t even be a possibility but now it is very plausible. I think it shows a growing crack down on free speech and our rights. Bigger picture to me is that we are going to be unjustly held accountable for things that are much within our right to do/possess. 

What would you say to the Norwegian guy if you could?

MCNAMEE: I would probably say "my bad" and ask what it's like being named Mads Mikkelsen. 

Do you have a favorite Vance edit?

MCNAMEE: My favorite Vance Edit is probably the one someone did of him as the little boy from Shrek 2 with the giant lollipop...I didn't make that one but it uses the face of one of the edits I did and it is solid gold. 

I would like to add that this meme seems to have become the biggest meme of the 2nd Trump administration and one of the biggest political memes of all time and if it does enter a history book down the line I would like them to use a flattering photo of me.

This article was produced in collaboration with Court Watch, an independent outlet that unearths overlooked court records. Subscribe to them here.

This article contains references to sexual assault.

An Ohio man made pornographic deepfake videos of at least 10 people he was stalking and harassing, and sent the AI-generated imagery to the victims’ family and coworkers, according to a newly filed court record written by an FBI Special Agent.

On Monday, Special Agent Josh Saltar filed an affidavit in support of a criminal complaint to arrest James Strahler II, 37, and accused him of cyberstalking, sextortion, telecommunications harassment, production of a “morphed image” of child pornography, and transportation of obscene material. 

As Ohio news outlet The Columbus Dispatch notes, several of these allegations occurred while he was on pre-trial release for related cases in municipal court, including leaving a voicemail with one of the victims where he threatened to rape them.

The court document details dozens of text messages and voicemails Strahler allegedly sent to at least 10 victims that prosecutors have identified, including threats of blackmail using AI generated images of themselves having sex with their relatives. In January, one of the victims called the police after Strahler sent a barrage of messages and imagery to her and her mother from a variety of unknown numbers.

She told police some of the photos sent to her and her mother “depicted her own body,” and that the images of her nude “were both images she was familiar with and ones that she never knew had been taken that depicted her using the toilet and changing her clothes,” the court document says. She also “indicated the content she was sent utilized her face morphed onto nude bodies in what appeared to be AI generated pornography which depicted her engaged in sex acts with various males, including her own father.” 

In April, that victim called the police again because Strahler allegedly started sending her images again from unknown numbers. “Some of the images were real images of [her] nude body and some were of [her] face imposed on pornographic images and engaged in sex acts,” the document says. 

Around April 21, 2025, police seized Strahler’s phone and told him “once again” to stop contacting the initial victim, her family, and her coworkers, according to the court documents. The same day, the first victim allegedly received more harassing messages from him from different phone numbers. He was arrested, posted $50,000 bail, and released the next day, the Dispatch reported.

Phone searches also indicated he’d been harassing two other women—ex-girlfriends—and their mothers. “Strahler found contact information and pictures from social media of their mothers and created sexual AI media of their daughters and themselves and sent it to them,” the court document says. “He requested nude images in exchange for the images to stop and told them he would continue to send the images to friends and family.” 

The document goes into gruesome detail about what authorities found when they searched his devices. Authorities say Strahler had been posing as the first victim and uploading nude AI generated photos of her to porn sites. He allegedly uploaded images and videos to Motherless.com, a site that describes itself as “a moral free file host where anything legal is hosted forever!”

Strahler also searched for sexually violent content, the affidavit claims, and possessed “an image saved of a naked female laying on the ground with a noose around her neck and [the first victim’s] face placed onto it,” the document says. His phone also had “numerous victims’ names and identifiers listed in the search terms as well as information about their high schools, bank accounts, and various searches of their names with the words ‘raped,’ ‘naked,’ and ‘porn’ listed afterwards,” the affidavit added.

‘What Was She Supposed to Report?:’ Police Report Shows How a High School Deepfake Nightmare Unfolded

An in-depth police report obtained by 404 Media shows how a school, and then the police, investigated a wave of AI-powered “nudify” apps in a high school.

404 MediaJason Koebler

They also found Strahler’s search history included the names of several of the victims and multiple noteworthy terms, including “Delete apple account,” “menacing by stalking charge,” several terms related to rape, incest, and “tube” (as in porn tube site). He also searched for “Clothes off io” and “Undress ai,” the document says. ClothOff is a website and app for making nonconsensual deepfake imagery, and Undress is a popular name for many different apps that use AI to generate nude images from photos. We’ve frequently covered “undress” or “nudify” apps and their presence in app stores and in advertising online; the apps are extremely widespread and easy to find and use, even for school children.

Other terms Strahler searched included “ai that makes porn,” “undress anyone,” “ai porn makers using own pictures,” “best undress app,” and “pay for ai porn,” the document says. 

He also searched extensively for sexual abuse material of minors, and used photographs of one of the victim's children and placed them onto adult bodies, according to court records.  

The Delaware County Sheriff’s Office arrested Strahler at his workplace on June 12. A federal judge ordered that Strahler was to remain in custody pending future federal court hearings.

Fansly, a popular platform where independent creators—many of whom are making adult content—sell access to images and videos to subscribers and fans, announced sweeping changes to its terms of service on Monday, including effectively banning furries.

The changes blame payment processors for classifying “some anthropomorphic content as simulated bestiality.” Most people in the furry fandom condemn bestiality and anything resembling it, but payment processors—which have increasingly dictated strict rules for adult sexual content for years—seemingly don’t know the difference and are making it creators’ problem.

The changes include new policies that ban chatbots or image generators that respond to user prompts, content featuring alcohol, cannabis or “other intoxicating substances,” and selling access to Snapchat content or other social media platforms if it violates their terms of service. 

This article was produced in collaboration with Court Watch, an independent outlet that unearths overlooked court records. Subscribe to them here.

A family in Utah is suing the Republican National Convention for sending unhinged text messages soliciting donations to Donald Trump’s campaign and continuing to text even after they tried to unsubscribe.

“From Trump: ALL HELL JUST BROKE LOOSE! I WAS CONVICTED IN A RIGGED TRIAL!” one example text message in the complaint says. “I need you to read this NOW” followed by a link to a donation page.

The complaint, seeking to become a class-action lawsuit and brought by Utah residents Samantha and Cari Johnson, claims that the RNC, through the affiliated small-donations platform WinRed, violates the Utah Telephone and Facsimile Solicitation Act because the law states “[a] telephone solicitor may not make or cause to be made a telephone solicitation to a person who has informed the telephone solicitor, either in writing or orally, that the person does not wish to receive a telephone call from the telephone solicitor.”

The Johnsons claim that the RNC sent Samantha 17 messages from 16 different phone numbers, nine of the messages after she demanded the messages stop 12 times. Cari received 27 messages from 25 numbers, they claim, and she sent 20 stop requests. The National Republican Senatorial Committee, National Republican Congressional Committee, and Congressional Leadership Fund also sent a slew of texts and similarly didn’t stop after multiple requests, the complaint says. 

On its website, WinRed says it’s an “online fundraising platform supported by a united front of the Trump campaign, RNC, NRSC, and NRCC.” 

“Defendants’ conduct is not accidental. They knowingly disregard stop requests and purposefully use different phone numbers to make it impossible to block new messages,” the complaint says.

The complaint also cites posts other people have made on X.com complaining about WinRed’s texts. A quick search for WinRed on X today shows many more people complaining about the same issues. 

“I’m seriously considering filing a class action lawsuit against @WINRED. The sheer amount of campaign txts I receive is astounding,” one person wrote on X. “I’ve unsubscribed from probably thousands of campaign texts to no avail. The scam is, if you call Winred, they say it’s campaign initiated. Call campaign, they say it’s Winred initiated. I can’t be the only one!”

Last month, Democrats on the House Judiciary, Oversight and Administration Committees asked the Treasury Department to provide evidence of “suspicious transactions connected to a wide range of Republican and President Donald Trump-aligned fundraising platforms” including WinRed, Politico reported.   

In June 2024, a day after an assassination attempt on Trump during a rally in Pennsylvania, WinRed changed its landing page to all-black with the Trump campaign logo and a black-and-white photograph of Trump raising his fist with blood on his face. “I am Donald J. Trump,” text on the page said. “FEAR NOT! I will always love you for supporting me.”

CNN investigated campaign donation text messaging schemes including WinRed in 2024, and found that the elderly were especially vulnerable to the inflammatory, constant messaging from politicians through text messages begging for donations. And Al Jazeera uncovered FEC records showing people were repeatedly overcharged by WinRed, with one person the outlet spoke to claiming he was charged almost $90,000 across six different credit cards despite thinking he’d only donated small amounts occasionally. “Every single text link goes to WinRed, has the option to ‘repeat your donation’ automatically selected, and uses shady tactics and lies to trick you into clicking on the link,” another donor told Al Jazeera in 2024. “Let’s just say I’m very upset with WinRed. In my view, they are deceitful money-grabbing liars.” 

And in 2020, a class action lawsuit against WinRed made similar claims, but was later dismissed.

A survey of 7,000 Facebook, Instagram, and Threads users found that most people feel less safe on Meta’s platforms since CEO Mark Zuckerberg abandoned fact-checking in January.

The report, written by Jenna Sherman at UltraViolet, Ana Clara-Toledo at All Out, and Leanna Garfield at GLAAD, surveyed people who belong to what Meta refers to as “protected characteristic groups,” which include “people targeted based on their race, ethnicity, national origin, disability, religious affiliation, caste, sexual orientation, sex, gender identity, or serious disease,” the report says. The average age of respondents was 50 years, and the survey asked them to respond to questions including “How well do you feel Meta’s new policy changes protect you and all users from being exposed to or targeted by harmful content?” and “Have you been the target of any form of harmful content on any Meta platform since January 2025?” 

One in six of respondents reported being targeted with gender-based or sexual violence on Meta platforms, and 66 percent of respondents said they’ve witnessed harmful content on Meta platforms. The survey defined harmful content as “content that involves direct attacks against people based on a protected characteristic.”  

Almost all of the users surveyed—more than 90 percent—said they’re concerned about increasing harmful content, and feel less protected from being exposed to or targeted by harmful content on Meta’s platforms.

“I have seen an extremely large influx of hate speech directed towards many different marginalized groups since Jan. 2025,” one user wrote in the comments section of the survey. “I have also noted a large increase in ‘fake pages’ generating false stories to invoke an emotional response from people who are clearly against many marginalized groups since Jan. 2025.”

“I rarely see friends’ posts [now], I am exposed to obscene faked sexual images in the opening boxes, I am battered with commercial ads for products that are crap,” another wrote, adding that they were moving to Bluesky and Substack for “less gross posts.”

404 Media has extensively reported on the kinds of gruesome slop these users are referring to. Meta’s platforms allow AI-generated spam schemes to run rampant, at the expense of human-made, quality content. 

In January, employees at Meta told 404 Media in interviews and demonstrated with leaked internal conversations that people working there were furious about the changes. A member of the public policy team said in Meta’s internal workspace that the changes to the Hateful Conduct policy—to allow users to call gay people “mentally ill” and immigrants “trash,” for example—was simply an effort to “undo mission creep.” “Reaffirming our core value of free expression means that we might see content on our platforms that people find offensive … yesterday’s changes not only open up conversation about these subjects, but allow for counterspeech on what matters to users,” the policy person said in a thread addressing angry Meta employees.

Zuckerberg has increasingly chosen to pander to the Trump administration through public support and moderation slackening on his platforms. In the January announcement, he promised to “get rid of a bunch of restrictions on topics like immigration and gender that are just out of touch with mainstream discourse.” In practice, according to leaked internal documents, that meant allowing violent hate speech on his platforms, including sexism, racism, and bigotry.

Several respondents to the survey wrote that the changes have resulted in a hostile social media environment. “I was told that as a woman I should be ‘properly fucked by a real man’ to ‘fix my head’ regarding gender equality and LGBT+ rights,” one said.“I’ve been told women should know their place if we want to support America. I’ve been sent DMs requesting contact based on my appearance. I’ve been primarily stalked due to my political orientation,” another wrote. Studies show that rampant hate speech online can predict real-world violence.

The authors of the report wrote that they want to see Meta hire an independent third-party to “formally analyze changes in harmful content facilitated by the policy changes” made in January, and for the social media giant to bring back the moderation standards that were in place before then. But all signs point to Zuckerberg not just liking the content on his site that makes it worse, but ignoring the issue completely to build more harmful chatbots and spend billions of dollars on a “superintelligence” project.