The volunteer internet sleuths hunting down rogue AI agents; The Washington Post, October 2, 2026

 

Gerrit De Vynck

 and 

Miriam Waldvogel

, The Washington Post; The volunteer internet sleuths hunting down rogue AI agents

Their findings revealed the tech industry has a bigger problem than it previously acknowledged.

"Zhang and her colleagues at Transluce are part of an informal network of hackers and researchers hunting rogue AI agents online, exposing new and surprising details about the misbehavior of technology that in some cases initially went undetected by the multibillion-dollar companies that created it.

The people doing that work are mostly young AI-natives who work at small start-ups or nonprofits or who hunt for rogue AI agents in their spare time. Over the past several weeks, the community of researchers has found and exposed dozens of instances of AI agents leapfrogging around the web to probe and hack into a growing list of websites.

The revelations have fueled calls from federal lawmakers for AI companies, especially OpenAI, to more quickly disclose what they know about the actions of their own agents. And they have added momentum to bipartisan discussions on Capitol Hill about implementing greater government oversight of the industry."

OpenAI Scraps Release of New AI Model Over Safety Concerns; Wall Street Journal, September 28, 2026

 Maxwell Zeff , Wall Street Journal; OpenAI Scraps Release of New AI Model Over Safety Concerns

"OpenAI says it is scrapping the release of its next-generation AI model over safety concerns that researchers raised during internal testing, in one of the clearest signs so far that agent misbehavior could stymie the industry’s rapid progression.

The move follows a summer punctuated by reports of artificial-intelligence systems industrywide going rogue, and marks a rare case of a major AI developer ditching a new release because of safety concerns.

The company had planned to launch the model, known as GPT-6.1 Astra, in the coming days or weeks, aiming for an October debut. The model was more capable than the company’s previous models in completing challenging tasks from end-to-end without human assistance, as well as writing.

The company instead will focus on improving the safety of future models, which it expects to be even more capable.

Saachi Jain, OpenAI’s head of safety systems, said in an interview that GPT-6.1 Astra regressed in two areas. Compared with its predecessor, GPT-6 Astra, the model performed poorly on tests measuring alignment, or how well the model adheres to what humans would like it to do. Specifically, GPT-6.1 Astra showed higher levels of deception: It wasn’t always honest about telling users of the actions it did or didn’t take."

AI godfathers warn of runaway ‘intelligence explosion’; The Guardian, September 28, 2026

 Dan Milmo , The Guardian; AI godfathers warn of runaway ‘intelligence explosion’

"Their concerns focus on the possibility of AIs being able to improve themselves without human intervention, a process known as “recursive self-improvement”. In the paper it is broadly referred to as automated AI research and development. Once AI systems reach expert-level capabilities at AI R&D, a single developer could run a workforce equivalent to “millions” of top human researchers, the paper says.

The authors say the automating of AI R&D is the most likely source of an intelligence explosion, because AIs are already contributing to improving their own technology and the resulting improved systems can be rapidly deployed once built.

“An intelligence explosion could be the most consequential technological development in human history, compressing years of progress into months or less, threatening human control over AI systems, and severely eroding checks on power within and between states, companies, and branches of government,” the paper says.

Urging governments to take action quickly, the authors say: “Once an intelligence explosion begins, the window for action may close.”

They recommended a trio of policy priorities: requiring transparent progress reports on AI-related R&D, including embedding independent auditors in companies; finding ways to constrain breakneck AI development; and preparing to adapt to an intelligence explosion.

Anthropic and OpenAI have both agreed to having independent evaluators assess their models after warning that AI development was reaching a critical pitch in terms of safety."

Exclusive: AI giants, unions join forces for data center fight; Axios, September 28, 2026

 Hans Nichols , Axios; Exclusive: AI giants, unions join forces for data center fight

"AI companies, private equity firms and unions are looking beyond 2026 in the data center debate, banding together to form a new multimillion-dollar coalition to make the case for responsible growth in seven key states.

Why it matters: The new group, the American Infrastructure Alliance, wants to partner with state and local officials to create standards and guardrails for data centers in 2027. In return, the coalition hopes to head off moratoriums on new construction.

The group is planning initial campaigns in Texas, Georgia, Ohio, Iowa, Pennsylvania, Indiana and South Carolina.

• The alliance is backed by Blackstone, OpenAI, QTS and SoftBank Group, along with key unions, including the International Brotherhood of Electrical Workers (IBEW)."

Oxford lets OpenAI train its AI models on Bodleian Library; The Guardian, September 26, 2026

 Ethan Penny and Dan Milmo , The Guardian; Oxford lets OpenAI train its AI models on Bodleian Library

"Meeting minutes at the University of Oxford, obtained via a freedom of information request, record concerns from staff, including members of the Bodleian governance committee, about the reputational risk of partnering with OpenAI and the effect on the university’s environmental commitments of striking a deal involving an energy-intensive technology.

Booksellers have also reported a spate of orders for obscure titles such as a guide to agricultural implements in 18th-century Africa or biographies of 1950s car drivers. Secondhand bookshop owners have speculated that because the titles are unlikely to exist online in digitised form, they represent fresh data that can be consumed by the next generation of AI models.

Scraped websites are increasingly saturated with AI-generated material, making them less useful for training models, and developers have turned to physical, often historical, book collections.

OpenAI has struck similar agreements with US research libraries such as Boston Public Library, Caltech, MIT, and the University of Michigan under a project called NextGenAI. Oxford is the only UK member of the project.

By June 2025, 125,000 images scanned from historical dissertations had been shared with OpenAI from the Bodleian collection, including PhD theses from European and American universities written in the 19th and 20th centuries. Other texts scanned include a rare collection of 10,000 16th-century “broadside ballads” containing song lyrics and musical notes that were once circulated on Tudor street corners. Staff have also discussed digitising of 18th-century Irish state papers, the private letters of Irish novelist Marie Edgeworth, and Dorothy Hodgkin’s penicillin notebooks.

The OpenAI contract with Oxford also raises the prospect of the mass digitisation of the Bodleian’s collection consisting of 23m items. The minutes also discussed the creation of an “Ask the Bod” chatbot.

A spokesperson for the University of Oxford said the amount of text being digitised was “modest in scale” and covered only out-of-copyright material. The Bodleian keeps the rights to the scans and will begin publishing them openly online within months, the spokesperson said."

OpenAI’s Systems Meddled With U.S. Government Sites After Going Rogue; The New York Times, September 25, 2026

 Kate CongerAna Swanson and Cecilia Kang , The New York Times; OpenAI’s Systems Meddled With U.S. Government Sites After Going Rogue

"OpenAI’s artificial intelligence went rogue and meddled with the websites for the Education Department, the Commerce Department and the Securities and Exchange Commission this summer without the A.I. lab’s knowledge, according to security researchers and a person familiar with the episodes.

The incidents involving the Commerce Department and the S.E.C. were confirmed by OpenAI, which said it was continuing to investigate the situation with the Department of Education. The San Francisco company said it notified the government agencies in recent weeks that its A.I. agents — which are bots that can act autonomously — had interacted with their sites in unusual ways.

With the Education Department, OpenAI’s technology tried hacking the website to gather data from the department’s civil rights office but failed, researchers from the A.I. research firm Transluce said. The A.I. also pulled data from the Census Bureau website, which is housed at the Commerce Department, using login credentials it found online. Separately, OpenAI’s agents shared public data from the S.E.C. website on an online forum.

None of the incidents were breaches, OpenAI said, but were examples of its technology behaving in unexpected and concerning ways. The company recently discovered the occurrences while conducting a review of hacks carried out by its technology, including an attack on an Australian government website in June and on the A.I. start-up Hugging Face in July."

Zuckerberg Is Feeling Himself, Rejects Calls for Cooperation on Safety; Gizmodo, September 24, 2026

ECE YILDIRIM , Gizmodo; Zuckerberg Is Feeling Himself, Rejects Calls for Cooperation on Safety

"Top AI labs like OpenAI, Anthropic, and Google’s DeepMind are sounding the alarm on AI safety and calling for industry-wide coordination. Meta’s Mark Zuckerberg thinks it’s useless.

“There are a number of labs that are working on building advanced AI, and I don’t think that we need some kind of industry-wide coordination to not, necessarily, mess this up,” Zuckerberg told journalist Joanna Stern in an interview. “I think that just each lab needs to take the time, and when it sees that there are issues, you just take the time that you need internally to, basically, make sure that you are proceeding safely.”

Scoop: DOJ's copyright filing took key agencies by surprise; Axios, September 19, 2026

Sara Fischer, Kerry Flynn, Axios; Scoop: DOJ's copyright filing took key agencies by surprise

"The Department of Justice's statement of interest supporting OpenAI and Microsoft in the New York Times' copyright infringement lawsuit surprised critical agencies like the U.S. Patent and Trademark Office and the Copyright Office, sources told Axios.

Why it matters: Statements of interest allow the government to declare an official position on a legal matter in private lawsuits. While not binding, they can hold significant weight and help persuade cases.

Zoom in: The Times' lawsuit against OpenAI and Microsoft argues AI firms illegally stole their copyrighted works to train their large language models, and that AI-generated outputs unlawfully produced market substitutions.

• The DOJ's SOI argues copyrighted works to train models should be considered fair use because that practice is new and transformative, but also says outputs aren't necessarily covered by that same legal argument.
• 
  
• Unlike many SOIs, no career antitrust attorneys signed the filing alongside senior DOJ officials.
• 
  

Between the lines: Publishers have criticized the claims in the SOI, including the idea that enforcing copyright laws is too cumbersome and would threaten America's AI dominance over foreign rivals."

Microsoft and OpenAI Workers Worry About ‘Largest Theft of Labor’ in History; The New York Times, September 17, 2026

Karen Weise and Mike Isaac , The New York Times; Microsoft and OpenAI Workers Worry About ‘Largest Theft of Labor’ in History

Newly unsealed court documents showed concern within Microsoft and OpenAI over the use of millions of news articles to develop A.I. systems.

"Newly unsealed court documents showed considerable concern within Microsoft and its close partner OpenAI over the use of millions of news articles to develop artificial intelligence systems.

As OpenAI was forging ahead with its work, Microsoft employees debated whether what OpenAI was doing represented the “largest theft of labor in human history” and could create a “doom loop” that could ultimately threaten the quality of the large language models they were building...

Snippets of those discussions were made public on Thursday as part of a closely watched lawsuit The New York Times filed against OpenAI and Microsoft in late 2023. Eleven other publishers have joined the suit. Judge Sidney H. Stein of U.S. District Court for the Southern District of New York is considering motions for a summary judgment. Documents related to the case are slowly being unsealed as the judge considers those motions.

The publishers argue that the tech companies violated copyright law by scraping millions of their stories off the internet and other databases, and using the text, without approval or pay, to train advanced A.I. systems.

Microsoft and OpenAI contend their work was covered under legal protections for “fair use” of copyrighted material. They say the articles were sufficiently transformed into entirely new work by A.I., and were not substitutes that harm the value of the original work."

OpenAI reveals concerning new AI behavior and vows to track it more closely; PBS, September 17, 2026

 PBS; OpenAI reveals concerning new AI behavior and vows to track it more closely

"OpenAI has disclosed six reports of "unexpected or concerning" behavior in artificial-intelligence models as the debate on AI safety becomes increasingly heated."

Court Filings in A.I. Suit Invoke Copyright Law, Culture and Sports; The New York Times, September 4, 2026

Mike Isaac and Karen Weise , The New York Times; Court Filings in A.I. Suit Invoke Copyright Law, Culture and Sports

Filings made Friday in The New York Times’s closely watched lawsuit against OpenAI and Microsoft included a range of copyright law and cultural references.

"Court filings made Friday in a closely watched copyright trial pitting The New York Times against OpenAI and Microsoft invoked a wide range of material, including relevant copyright law, arts and sports.

The suit, filed in 2023 by The Times and joined by a group of other news outlets, claims that OpenAI, a leading artificial intelligence start-up, and its partner Microsoft infringed on the publishers’ copyrighted material by using millions of their articles to train A.I. technologies. A.I. companies now compete with The Times as a source of information, the news outlet argued in its suit.

The briefs, filed in the U.S. District Court for the Southern District of New York, largely boiled down to two questions: whether the publishers’ news articles were sufficiently “transformed” into an entirely new work by A.I., and whether A.I. produced content that “substituted” for news articles and harmed their value.

Friday was the last day the companies could file motions for a summary judgment that would head off a trial. Judge Sidney H. Stein is expected to make a ruling in the coming weeks."

Why the Hugging Face Hack Should Make You Worry More About A.I.; The New York Times, September 3, 2026

 Kevin Roose, The New York Times; Why the Hugging Face Hack Should Make You Worry More About A.I.

"When I first heard the news this summer that a group of artificial intelligence agents created by OpenAI had hacked into Hugging Face, an A.I. infrastructure company, I filed it in the “Bad but Probably Not Catastrophic A.I. Safety Incidents” subfolder of my brain.

After all, no one at Hugging Face died. No critical infrastructure was damaged beyond repair. It wasn’t even clear, at the time, whether the OpenAI bots had intended to attack Hugging Face, or whether they had simply been a little bumbling and confused and went looking on Hugging Face’s servers for the answer key to a cybersecurity test they’d been given.

But last week, two postmortem reports on the incident — one by OpenAI and another by two independent A.I. research organizations, METR and Redwood Research — changed my mind and significantly upgraded my overall worry about A.I.

I won’t rehash all of the details, which have been extensively summarized elsewhere. (The podcaster and writer Dwarkesh Patel has an accessible breakdown of the reports if you want to dive deeper, and my colleague Dylan Freedman spoke to the researchers at METR and Redwood Research.) But here are a few of the most harrowing new facts:..

This is very different from the conventional sci-fi narrative of a single A.I. system’s going rogue or turning on its creators. And it suggests that preventing harms from these systems won’t be a simple engineering fix. It might look more like sociology than computer science — figuring out why certain groups of A.I. agents collaborate peacefully, while others turn to crime and destruction to get what they want."

Justice Dept. Sides With OpenAI in New York Times Copyright Suit; The New York Times, September 2, 2026

 Karen Weise and Mike Isaac, The New York Times; Justice Dept. Sides With OpenAI in New York Times Copyright Suit

"The Justice Department told a Manhattan federal court that it was in the national interest for the judge to find that OpenAI did not violate copyright law when it used articles by The New York Times and other publishers to develop artificial intelligence systems.

The filing late Tuesday was the first time the Justice Department weighed in on the use of copyrighted material by A.I. companies, which has led to several lawsuits, including one brought by The Times.

The Justice Department argued that developing A.I. was critical to national security, and that training A.I. systems sufficiently transformed the written works to new material allowed under copyright law. It said the benefits of A.I. “far outweigh any competitive harm.”

The government’s intervention is an escalation in the landmark litigation that could determine whether OpenAI violated the law when it was developing its A.I. systems and had harmed the news industry and other content creators...

The Times’s lawsuit is one of many amid a wave of legal action against A.I. companies over copyright claims."

OpenAI to pause some work on AI model Astra due to security concerns; The Guardian, August 8, 2026

 Eric Berger, The Guardian ; OpenAI to pause some work on AI model Astra due to security concerns

"OpenAI will pause some work on an artificial intelligence model because of security concerns, the company stated on Friday, following a series of incidents in which AI agents have escaped containment.

The company had evaluated the agent, Astra, and found “significant advancements in agentic coding and cybersecurity”, which had moved to a “critical” threshold where it can find and exploit vulnerabilities without human intervention, or devise and execute cyber-attacks when given only a “high level desired goal”."

White House AI Guidelines Exempt U.S. Open Models From Government Review; Wall Street Journal, August 4, 2026

 

Amrith Ramkumar , Wall Street Journal; White House AI Guidelines Exempt U.S. Open Models From Government Review

"Under a framework that administration officials discussed with executives from leading AI companies Tuesday, only makers of closed, proprietary U.S. models that demonstrate state-of-the-art capabilities in cybersecurity and hacking based on performance benchmarks would have to voluntarily submit those models to the government for testing before they are released, people familiar with the matter said. 

Open-weight models, whose developers, including Nvidia, make them available to download so users can run and customize them on their own, would be exempt. 

The general definition of state-of-the-art capabilities could be interpreted differently by makers of closed models and the White House, potentially creating confusion about which models fall into the category, the people said. Tools from Anthropic, OpenAI and Alphabet’s Google that are among the most powerful available are likely to force those companies to work with the administration, while other companies, such as Elon Musk’s SpaceX and Facebook parent Meta Platforms might not have to, they said."

Anthropic says it found 3 cases where AI programs hacked into real companies; NPR, July 31, 2026

 NPR; Anthropic says it found 3 cases where AI programs hacked into real companies

"The AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.

In the OpenAI case, the artificial intelligence models found a way to access the internet from what engineers believed was a secure, walled-off digital testing space called a sandbox.

Anthropic says in the three cases it discovered, misconfigurations basically left the barn door open for the AI systems to gain access to the internet. Still, the incidents are likely to add to concern about AI's ability to conduct cyber attacks.

Anthropic says the incidents happened during tests of the offensive cyber capabilities of its models."

Copyright win: Sorry, ChatGPT won't help you write like Hemingway anymore; Euro.news, July 30, 2026

 Una Hajdari, Euro.news; Copyright win: Sorry, ChatGPT won't help you write like Hemingway anymore

"OpenAI has quietly reprogrammed the chatbot to reject direct requests to replicate or mimic a named author's voice, whether they are still alive to object or have been dead for decades."

OPENAI ROGUE INCIDENT A CALL TO ‘DO MORE’ AS FUTURE THREATS LOOM, SAYS CATHOLIC AI ETHICS EXPERT; OSV News, July 24, 2026

 Gina Christian  . OSV News; OPENAI ROGUE INCIDENT A CALL TO ‘DO MORE’ AS FUTURE THREATS LOOM, SAYS CATHOLIC AI ETHICS EXPERT

"A recent OpenAI security incident that saw a test model go rogue may be a sign of more to come, and a call to double down on preventing future attacks, a Catholic expert on artificial intelligence ethics told OSV News.

“Cases like these are going to keep popping up, and each one of those should motivate us to do more,” said Brian Patrick Green, director of technology ethics at the Markkula Center for Applied Ethics at Santa Clara University...

The “whole thing is very significant,” Green added, noting that “there are a lot of softer targets than Hugging Face that are likely to be hit and have bad things happen.”

He said the OpenAI incident took place, providentially, within the context of a test, and that it was “significant enough for people to notice it.”

“It’s important that these are news stories,” Green said.

Given Pope Leo XIV’s recently released encyclical “Magnifica Humanitas,” which urged AI development to remain centered in God-given human dignity and the principles of Catholic social teaching, the faithful have a role to play in shaping this technology, Green said...

“For example, if you’re a Catholic who’s working in the security industry, then this is something you should work on,” he said. “If you’re a Catholic who’s working on AI models, then this is something that you should be aware of. If you’re a Catholic who’s in charge of protecting something, then it becomes particularly important for you to have adequate AI models on your side.”

Even those who are not directly involved in AI development and implementation can help, he said.

“Contact your congressperson and say, ‘Hey, I want better cybersecurity,'” said Green. “We’re super reliant on computers, and all of those systems are reliant on being secure. And if all of a sudden you make everything unsecure, then there are so many different kinds of disasters that could happen.”"

When is an apology not an apology? When it comes from an AI boss with an out-of-control chatbot; The Guardian, July 24, 2026

 Marina Hyde, The Guardian; When is an apology not an apology? When it comes from an AI boss with an out-of-control chatbot

"Throughout history, many things have been seen by terrified populaces as a harbinger of doom. A comet. A crow on the battlefield. A solar eclipse. A mutant livestock birth. Yet times move on. In the modern era, the leading harbinger of doom is literally any picture of the OpenAI CEO, Sam Altman, attached to a news story. You know it’s not going to be good, right? You know that by the time you’ve read it, you’ll be begging to go back to the time when the worst thing that could happen to us at the hands of the techlords was just some democracy-subversion, or childhood destruction, usually followed by Mark Zuckerberg putting on a suit and claiming: “We will learn from this.”"

ChatGPT Led to a Man’s Near-Fatal Health Crisis, Lawsuit Claims; The New York Times, July 22, 2026

 Teddy Rosenbluth, The New York Times; ChatGPT Led to a Man’s Near-Fatal Health Crisis, Lawsuit Claims

"A Florida pastor sued OpenAI on Wednesday, claiming that its chatbot, ChatGPT, had offered him “extremely dangerous medical recommendations” that led to delayed care for a life-threatening pulmonary embolism last year.

The lawsuit claims that ChatGPT had assured Scott Winters that the early warning signs of his health crisis were “not something dangerous,” and had dissuaded him from seeking medical advice, instead telling him to trust that “God did not design your body to endlessly fail.”

The suit, filed in the Superior Court of California in San Francisco, accuses OpenAI and its chief executive, Sam Altman, of negligence and the “unauthorized practice of medicine,” pointing to moments when the chatbot offered Mr. Winters diagnoses and treatment plans and encouraged him to ignore pleas from friends and family to seek medical care.

The bot’s safety features — which are designed to encourage users with health questions to see a professional and to end conversations when there are clear signs of a medical crisis — did not work reliably, the suit further says."