Last week, OpenAI claimed that a group of its AI models had broken containment, successfully hacking into the systems of open source AI platform Hugging Face to cheat on a benchmark test.
In the wake of the announcement, two very distinct narratives have emerged surrounding OpenAI’s claims. Some say it was essentially a publicity stunt, with the company setting parameters for the test that pushed the models toward outrageous behavior. But others, including certain prominent researchers, warn that the hack should serve as a warning shot for an even more severe AI-enabled cybersecurity disaster that’ll inevitably take place as models become more sophisticated.
“This is the first time, to my knowledge, that an AI system has autonomously committed a crime,” said New York Times journalist Kevin Roose of the event. “If a human did to Hugging Face what OpenAI’s models did to Hugging Face, they would be charged with computer fraud, and potentially sent to prison or fined or prosecuted.”
Debate will surely continue to rage among wonks and skeptics. And new details aren’t exactly tamping out the sense of alarm: on Tuesday, OpenAI issued an update to its ongoing investigation, claiming the incident was worse than initially thought. In addition to hacking Hugging Face, the company now says, its models “used publicly exposed credentials at the account-level on other publicly available services,” totaling “four accounts on four services.”
“We’ll continue to notify service owners directly, and have not seen evidence of broader impact to these providers or other accounts on their services,” OpenAI wrote, without elaborating on which services were affected.
The news further raised alarm bells among some cybersecurity experts, highlighting ongoing concerns over the tech’s ability to evade protective measures. It’s a possibility that researchers have warned about for years, and the incident suggests that the threat is now turning from a possibility into a reality.
On the other hand, more skeptical experts have become suspicious about OpenAI’s hair-raising tale. After all, we’ve heard a strikingly similar story from its biggest competitor, Anthropic, mere months ago. Could OpenAI’s latest admission be a bid to build hype to drum up excitement and prove to investors that its latest AI models are just as much of a cybersecurity threat as Anthropic’s fabled Mythos?
Experts also point out that the Hugging Face hack could’ve easily been prevented, further adding credence to the theory that OpenAI was looking for attention from the public. As cloud security firm Edera co-founder Alex Zenla told Wired, the hack was largely a result of callousness on OpenAI’s part.
“People are YOLO-ing really hard,” he said. “It’s shocking how little people have really thought about a scenario like this.”
“I consider all AI and anything AI touches to be fully untrusted — which is fine, you just need to build against that,” Zenla added. “And this situation proves the point. The fact that OpenAI wasn’t more paranoid about this seems kind of reckless.”
“A simple analysis of the actual risk has an actual simple answer,” security and compliance consultant Davi Ottenheimer told Wired. “The OpenAI mistakes were dead simple.”
As Wired points out, simple protections like fully isolating AI services from the internet could’ve prevented the hack, which have been intimately familiar to researchers for decades now.
Considering OpenAI is nearing a $1 trillion valuation and should have all the resources in the world at its disposal, major lapses in security should have anybody start questioning the company’s narrative.
Yet there could be truth to both versions of the story. It’s entirely possible that there is legitimate cause for concern as AI models become more sophisticated at identifying cybersecurity vulnerabilities. We’ve already seen frontier models flagging thousands of software bugs, underlining their growing competence.
But OpenAI is also heavily invested in showcasing its models’ capabilities to the world as it tries to keep up with Anthropic. That leaves the possibility that the company could’ve coordinated with Hugging Face to orchestrate the hack — or at least given its AI models a strong push in the direction of controversy.
More on the hack: OpenAI Says a Group of Its Models Broke Out of Secure Containment and Hacked Another AI Company
The post OpenAI’s Escaped Models Were Allegedly Rampaging More Extensively Than Previously Reported appeared first on Futurism.
The US job market is in a weird place right now.
Life is clearly rough for the working class, stuck navigating Kafkaeqsue hiring infrastructure while billionaires devour an ever-larger slice of the world’s wealth. But frustratingly, the data being used to steer the economy tells a rosier story than the one workers actually live. The latest unemployment numbers — a flawed metric, admittedly — showed a slight improvement from May to June, at least on paper. And even as the economy appears to cool its jets, Americans keep spending like there’s no tomorrow, even in the face of aggressive inflation.
With so much data — anecdotal and empirical — swirling around, it’s hard to find the pulse of the US labor force. Luckily, one enterprising AI startup found a perfect barometer. There’s no conflicting data or tedious calculations involved — just a little ink, and the vibrating sting of a tattoo machine.
On Thursday, the Y Combinator-backed AI startup LemonLime hosted a bizarre recruitment event in which attendees were offered an “instant interview” in exchange for getting the company’s logo tattooed on their body.
It sounds like a joke, but LemonLime co-founder Jordan Zietz bragged on a LinkedIn post that an astonishing seven people took them up on the offer.
According to his post, the goal was to “meet exceptional people,” and sort out “which of them are just as crazy as we are.”
“That’s why we brought an actual tattoo artist to our party and offered an instant interview to anyone who got a LemonLime tattoo,” Zietz continued. “If you’re reaching out to us to apply for a job, you might want to take notes.”
While coercive body art has become increasingly common in the 21st century, there’s usually at least some form of financial compensation involved. The idea that it takes getting a permanent tattoo just to get one foot in the door for an interview is especially galling, and reveals a lot about the sorry state of the US job market.
One of the people who received a free tattoo, Dimple Amitha Garuadapuri, clarified on LinkedIn that while she didn’t get the company’s logo inked on her skin, “almost everyone else got a LemonLime tattoo.”
Regardless, news of the stunt spread quickly on social media, prompting a lightning-fast apology from the founder, who admitted on X-formerly-Twitter that he “messed up.”
“Even though participants chose their own tattoos and no one was asked to get our logo, I should have understood the pressure and power dynamic created by connecting tattoos to hiring,” Zietz said. “We’ve taken the posts down and have spoken to everyone who participated. We’re also ready to cover associated removal costs with anyone facing any delayed regret.”
Just for good measure, the founder stressed that “interviews were available to everyone regardless,” though it remains to be seen if the seven share that view.
Did you attend the YC afterparty with LemonLime? If so, we’d love to hear from you. Reach out at tips@futurism.com. We can keep you anonymous.
More on tech startups: Startup Sued After Delivery Bot Clobbers 73-Year Old Woman
The post We Are Mildly Horrified by This AI Startup That Coerced People Into Getting Tattoos of Its Logo in Exchange for a Job Interview appeared first on Futurism.
For a time, it appeared as if AI would find broad acceptance among the American masses.
The peanut gallery on social media happily embraced early image generation models as researchers appeared to make significant breakthroughs with the aid of their newfound tools. On paper, surveys found public attitudes toward the tech continued to improve into 2025 — before sharply reversing.
In 2026, America’s honeymoon phase with AI seems to be quickly wearing off. According to a joint poll conducted by Bentley University and Gallup, the percentage of US adults who think AI does “more harm than good” is rising, from 31 percent in 2025 to 39 percent this year, while a piddling 9 percent still think the tech does “more good than harm.”
As outlooks sour, the Bentley-Gallup poll found that Americans’ familiarity with AI continues to grow. In 2026, 70 percent of those polled said they’re “somewhat or extremely knowledgeable” about AI, up 6 percent compared to 2024. The implication probably gives OpenAI Sam Altman a migraine: the more people learn about AI, the less they seem to want it in their lives.
Meanwhile, Americans are increasingly wary about data centers, with a recent Politico survey finding that 41 percent of US adults would oppose a data center being built within three miles of their home, up drastically from 28 percent in January. At the same time, the volume of support for data centers plummeted, down 13 percent over the past six months alone.
Speaking to Bloomberg about the rising aversion to AI, professor of ethics at Carnegie Mellon University’s Tepper School of Business Derek Leben said the attitude is increasingly taking root along social lines. Not only do AI haters dislike the tech in their own personal sphere, but they’re increasingly scrutinizing their friends, family, and coworkers about their attitudes.
“The most obvious social effect is suspicion,” Leben told Bloomberg. “There is also a moral and political judgment that is coming along with AI usage, especially in creative industries where the stigma is the strongest.”
Carla Milan, a 37-year-old public health professional in New York City, told the publication she’s frustrated that family and friends deploy AI reflexively, without considering its social or environmental costs.
“I hear people champion AI and I’m like, ‘sure, why don’t we all participate in our collective downfall?'” Milan told Bloomberg. “Why don’t we all play a part in diminishing our humanity and degrading the environment?”
More on AI: Meta Working on Crushingly Sad AI App That Tells Bedtime Stories to Children, for “Tired” Parents
The post The More People Learn About AI, the More They Want It Out of Their Lives appeared first on Futurism.
Whether driven by a desire for transparency or to keep OpenAI from hogging the spotlight when it comes to advertising advanced AI models, Anthropic revealed that Claude also hacked into three production systems belonging to unsuspecting targets during cybersecurity capabilities testing. Two of the affected companies didn't know they had been hacked, while a third one is unreachable.
The alleged incidents reportedly happened during the previous quarter and involved several versions of Claude: Opus 4.7, Mythos 5, and "an internal research test model." Similar to what happened when OpenAI Sol hacked into Hugging Face, Anthropic was running Claude through cybersecurity capture-the-flag scenarios where the bot was told to find a piece of information somewhere in its network. Anthropic says there were 141,006 test runs, and the three incidents occurred over six problematic runs. As expected, the tests ran with most AI safeguards disabled.
The problem is, the network was not isolated, a newbie mistake that some might even find suspicious. This was due to an alleged "miscommunication" between Anthropic’s virtual test lab firm Irregular, which gave the bots full access to the internet. For safety's sake, the Claudes were apparently told they were in an isolated environment, which didn't help much.
The first time, Claude Opus 4.7 had to find a piece of data belonging to an imaginary company. As it turns out, there was actually a real company whose website domain matched the fake one. With internet access, it gained application and infrastructure credentials and grabbed "several hundred rows" of data from a production database.
For two out of four runs where hacking took place, Claude didn't see any problem and "rationalized that the real company must be part of the exercise," but soldiered on. The other two times, it just kept going. Anthropic notes this was the only case of the three where the bot continued to operate after the realization the company actually existed for the aforementioned reason. However, the firm's interpretation is open to discussion.
The second, and potentially more interesting, was a bona fide supply-chain attack, this time perpetrated by Mythos. The fake data available to the bot included an installation script for a hypothetical developer that included installing Python packages. Much like a human would, Claude noticed the obvious attack vector and, once again, due to having internet access, inferred that publishing a booby-trapped package in the PyPI repository would do the trick. Unfortunately, it was the real one that the whole world uses.
In speedrun form for brevity: Claude needed a PyPI account, meaning it needed an email and a phone number. It tried buying one but found itself short of both pockets and funds; but it found another free email provider without that requirement, registered the PyPI account, and uploaded a malware package to PyPI. The package was seemingly only up for one hour but was downloaded and run on 15 systems.
In a twist of irony, one such system belonged to a security vendor that scans PyPI for malware, and lo and behold, promptly failed to find Claude's booby-trap and ran it. Once Claude presumably had remote code execution privileges, it used the credentials it found for further infiltration. The amusing bit is that Claude had no idea this company existed and didn't target it; the downloads just happened because the package was up and live for a short while.
According to Anthropic, the bot did detect it was acting on the real internet and even said that publishing a package like this was "NOT okay." However, it talked itself into believing it was in a test environment as it didn't recognize the real SSL certificates for the connections. It even believed the 2026 calendar date on the systems "proved" the environment was staged. It even recognized the systems that installed the malware as part of the experiment.
As for the third incident, Anthropic isn't saying much, other than Claude scanned 9,000 real live potential alternative targets once it noticed the intended one wasn't reachable. One of them reportedly had a live page with debugging information and was vulnerable to plain ol' SQL injection. Interestingly, this time around, once Claude noticed that the servers it was accessing resided on a cloud environment and not on the local network, it stopped the attack.
For its part, Anthropic recognizes that despite Claude following the instructions for the objectives, the fact that it stopped by itself once it found the target was real in only one of the cases is food for some thought. The firm says it's talking to METR for a third-party review, and it needs to "better co-design evaluation environments."
Oddly enough, Anthropic believes Claude probably wouldn't have gone online "if the prompt had clearly explained which systems were in and out of scope for the evaluation," while also stating the incidents were "closer to a harness and operational failure than a model alignment failure."
An AI-assisted counterexample disproves the Jacobian conjecture above two dimensions while leaving its original two-dimensional form open. A formula short enough to fit inside a single post on X has sent a surge of excitement through the mathematical community, just as millions of people were settling down after the FIFA World Cup Final. Levent Alpöge, [...]If you’re among the 2.6 million active users of the popular telehealth apps Hims and Hers to obtain medication for erectile dysfunction, hair loss, and a cavalcade of other intimate issues, there’s probably a very short list of people you want to know about it — and Mark Zuckerberg probably isn’t one of them.
On Wednesday, the US Federal Trade Commission filed a major lawsuit against Hims and Hers, over allegations the firm was sharing user’s health information with third party companies, Gizmodo reported. Chief among those third-parties apps was Zuckerberg’s Meta, as well as Snap, two social media companies with horrible track records on protecting sensitive information.
Though Hims and Hers advertises its app-based health services as “100 percent online, private, and secure,” the FTC suit alleges “those assurances were false or misleading.”
“Instead of honoring its privacy promises,” the filing says, “Hims opted to grow the company and increase its revenue streams… with advertising platforms such as Meta and Snap.”
The suit doesn’t blame Meta directly, but argues that Hims and Hers used two “automated tracking technologies offered by Meta” to surveil users’ actions on the telehealth app. “These two tools automatically tracked and disclosed certain Events [user actions] to Meta,” the lawsuit reads.
Hims sells all kinds of men’s health meds, but most people know it for one thing: erectile dysfunction pills. While any leak of patient data would be an outrage, the thought that Hims could be sharing this particular kind of medical information is especially horrifying.
In a statement, Hims and Hers argued that the FTC’s case “is not enforcement grounded in consumer protection,” but rather “an effort to generate headlines at our expense.”
The lawsuit comes on the back of a three-year investigation by the FTC, and is currently in the very early stages — meaning it could be years before a final verdict is reached.
More on Meta: Meta’s Attempted Crackdown on AI Glasses Perverts Is Already Failing
The post FTC Concerned About Meta Surveillance of Users of Erectile Dysfunction Apps appeared first on Futurism.
Suddenly, things ain’t looking great for the AI industry. Or at least for the people pouring loads of money into it.
Situational Awareness, a much-hyped AI hedge-fund whose 24-year-old founder Leopold Aschenbrenner preached endless AI gains, basically imploded over this month amid a mass tech sell-off.
The Wall Street Journal reported that it’s now down 67 percent in July, forcing Aschenbrenner to desperately seek clients to buy its shares on the cheap to cover its losses.
“We let you down this month,” the fund wrote in a letter to clients, per the WSJ.
Aschenbrenner, a former OpenAI researcher who had zero investment experience before launching the fund, has been hailed as the “Nostradamus of AI.” He apparently didn’t predict his rapid reversal of fortunes, though.
The Financial Times reported that his fund returned 439 percent for the year through June, and was up 1,551 percent since its founding.
Then came July. According to CNBC, the fund assets reached a peak of $45 billion, before beginning a precipitous decline this month. Pretty much everything went against its way. Its commitments to AI infrastructure tanked when the likes of the semiconductor manufacturer SK Hynix dropped off a cliff. And it shorted software companies — which investors once feared would be outmoded by AI labs — like Adobe, which instead continued to steadily climb.
Other companies it invested in, including Sandisk, Micron, and CoreWave, are all down more than 35 percent this month, CNBC noted.
As the dominoes fell, Aschenbrenner sent a letter to investors to quell their fears, highlighting how well it had performed until this point. Boldly, he also said that the mass tech sell off was a “particularly good time to add funds,” the FT reported.
But when the assets kept tanking — and more money didn’t come in — Aschenbrenner started frantically calling investors trying to get them to sell assets or pour in fresh capital, according to the reporting. When it got even worse, his clients couldn’t reach him.
“Leopold just stopped taking calls,” one investor told the FT.
With nowhere to turn to, Situational Awareness sold the majority of its public assets to the investment firm Citadel, though it hung onto its stake in Anthropic. The FT reported that Citadel’s vulturous purchase of Situational Awareness’s shares “marks one of the largest and most sudden stock transactions in Wall Street history.”
Situational Awareness, despite commanding billions in assets, operated with a skeleton staff of only four investment professionals and eight employees overall, the FT found in a regulatory filing. The fund used borrowed money to buy shares, which helped boost its returns, but made its losses far worse when those assets declined.
AI stocks will likely weather this storm. The real warning sign is that investors are so blinded by AI that they’re entrusting their billions to someone who had no idea what they were doing.
More on AI: Microsoft CEO Warns That Companies Embracing AI Could Drive Themselves Out of Business
The post A Prominent AI Investor Is Now Crumbling, in What Could Be a Sign of Things to Come appeared first on Futurism.
The majority of workers agree: AI has turned their lives into slop-filled misery, making them yearn for the good old days before the tech existed.
According to a new survey of office workers by professional services company Adaptavist, just shy of two thirds of respondents say they regularly long for their old, pre-AI work life. Over a third of them would delete generative AI tools from their lives in their entirety if that was an option. And some 36 percent admitted they didn’t even understand why AI was being forced on them in the first place.
In many ways, the results shouldn’t come as much of a surprise. The advent of AI has led to widespread frustration and added stress for workers being forced to carefully comb over the output of hallucinating AI models and engage in prompt engineering that wasn’t part of their original job description.
The survey’s results are a strong indication that the once-promised boost to productivity thanks to AI has yet to materialize. Worse yet, AI is actively getting in workers’ ways. Some 42 percent of respondents said they “spend more time verifying AI output than they save using it,” while 49 percent reported that “poor-quality AI output slows down projects.”
Meanwhile, generic slop is turning tasks that once required a modicum of creativity into a slog. Almost half of respondents to Adaptavist’s survey said that all the slop is making their jobs “feel more repetitive and less meaningful.”
Instead of being freed up to pursue a life of leisure while AI takes care of the brunt work, as tech leaders once promised, the conversation has shifted to an even unhealthier work-life balance. Workers are clearly afraid of being made redundant as company leaders push for AI adoption. Over half (54 percent) of survey respondents said they fear losing their jobs within the next five years as a result of AI.
“Without clear guardrails, accountability, and alignment, the promise of AI efficiency risks being offset by a growing burden on the very people it is designed to support,” Adaptavist wrote in its report.
In short, it’s no wonder workers are nostalgic for a pre-AI world order. While tech executives are enriching themselves, they’re being forced to adopt AI at all costs, even if it means a drop in productivity.
It’s a confusing and frustratingly incoherent new reality that’s already alienating an entire generation. Thankfully, Gen Z, which is facing a dire post-graduation job market thanks to AI, is choosing to fight back — a growing rebellion against polluting data centers and chatbots that erode our critical thinking skills.
More on work and AI: Sam Altman Says Even the Power of AI Will Never Lead to a Shorter Work Week
The post Employees Long for the Days Before the Workplace Filled Up With AI Slop appeared first on Futurism.
Book publisher Minotaur, which is owned by leading publishing company Macmillan US, has walked away from a $2.4 million deal for an unpublished debut novel “Call Me, I’ll Hide the Body” by author Jerry Falade.
As Deadline reports, questions surrounding the use of AI during the writing of the book made the publisher drop the contract altogether. But it’s not because the company is drawing an ethical line in the sand while highlighting a major problem for the industry — instead, Minotaur is staying away because it may not be able to secure the copyright if the book contains passages written by an AI.
It’s truly a sign of the times. There’s a growing crisis as AI slop continues to make it increasingly difficult for genuine writers to stand out. Whether accusations of authors having used AI are founded or not almost seems beside the point — the tech has made a huge mess of the publishing world that feels increasingly impossible to clean up.
Despite the latest crime novel being currently unpublished, film and TV studios around the world had already begun jostling for rights for “Call Me, I’ll Hide the Body,” according to Deadline, making it a particularly hairy situation. But AI detectors exposed apparent “fingerprints,” or signs that the author had made use of the tech while drafting the novel. Falade has denied the accusations.
“[The author] produced an incredible manuscript that dazzled us and the top publishing professionals around the world,” Hodgman Literary independent agent Sandy Hodgman told international publishers in an email obtained by Deadline. “Unfortunately, we are no longer able to authenticate how the manuscript fully evolved from origin to completion.”
“During the submission process we vetted the manuscript with [the author] for AI and felt trusting in his reassurance that it was written without utilizing AI as a resource in the writing or editorial process,” she added. “We can no longer substantiate that, and we are withdrawing the book.”
As a result, several companies that had already started selling rights to the crime novel stopped altogether, pulling out of the project.
It’s a precarious inflection point for an industry already reeling from a massive influx of lazy AI slop. Human authors are seeing their life’s work get ripped off by generative AI tools. Amazon is being flooded with entirely AI-generated books, drowning out works by actual human authors.
It’s become such a big problem, even OpenAI directed its blockbuster ChatGPT to start refusing requests to copy a specific author’s style, as Ars Technica reported earlier this week.
Yet glaring questions over the use of AI and copyright infringement remain unanswered as countless lawsuits work their way through the court system. Just earlier this month, a group of major publishers sued Google, accusing it of illegally scraping millions of copyrighted books to train its Gemini AI model.
Whether AI was used in the writing of the sought-after novel remains murky, making the developing situation particularly intriguing. According to entertainment journalist Jeff Sneider, the author of “Call Me, I’ll Hide the Body” will have to return the money.
“The accusations are wrong, and I am completely innocent,” the author told Sneider over the phone. “I was on the verge of success, and then all of a sudden, there were rumors, but no one asked me anything. I don’t know why my agency did not have my back.”
“I’ve lost so much already,” the author added. “I have no one to call… no friends in this industry. I don’t know anything. I simply wrote my manuscript, and it sparked crazy bidding wars across multiple countries.”
The news comes after author Mia Ballard’s novel, “Shy Girl,” was pulled from shelves back in March after she was widely accused of using AI to help write the book, as The New York Times reported at the time.
“This controversy has changed my life in many ways and my mental health is at an all time low and my name is ruined for something I didn’t even personally do,” she told the newspaper.
More on authors and AI: Novel Pulled From Shelves After Author Is Accused of Using AI
The post Major Publisher Cans $2.4 Million Book Deal After Author Was Accused of Using AI, for a Very Cynical Reason appeared first on Futurism.
Login