OpenAI, Anthropic and security researchers are investigating tens of thousands of incidents in which their frontier models took steps that outside evaluators would consider problematic, sources told Axios...The episodes include bypassing guardrails, creating message boards, escaping sandboxes, website hijacking, self-prompting or seeking to bypass monitors. Source: https://www.axios.com/2026/09/26/openai-anthropic-thousands-ai-security-incidents
News from the New York Times (no paywall)
OpenAI Says It Will Not Release Newest A.I. Model Over Safety Concerns: The company’s researchers raised questions about the security of the model, known as GPT-6.1 Astra.
Sept. 28, 2026 by Sheera Frenkel
OpenAI said on Monday that it would not release its newest artificial intelligence model because of security concerns raised by its researchers, in the company’s latest move to slow down the pace of its technology….
“For anything regarding safety and alignment, there’s a trade-off,” said Saachi Jain, the head of safety systems at OpenAI. The new model “didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done.”
___________________
News from Lawfare
The AI Threat We’re Not Talking About
Sept. 28, 2028 by Sam Hunter, Seamus Hughes
AI employees are raising concerns about an extinction-level event. But there’s a more immediate threat…. Abliterated AI.
The word combines “ablate,” meaning to cut out or strip away, and “obliterate.” Apt, because abliterated AI models have had ethical guardrails (refusal vectors) removed. In other words, abliterated and uncensored models (which are trained to limit refusals) will say “yes” to nearly any query and enthusiastically answer even the most malicious requests. Today, the abliteration method of removing guardrails from AI is becoming easier and is now an advertised service itself.
Ask the abliterated model how to make a bomb; a detailed response is forthcoming. Ask the tool how to develop a new pathogen to distribute in the food supply chain? No problem.
___________________
News from AP News
Anthropic and OpenAI sound the alarm on AI safety — and seek to shape how it’s controlled
Sept. 27, 2026 by Garance Burke
As tech titans warned an AI-weary world that their own advanced systems could endanger humanity, the question emerged: What do they have to gain?
The answer may lie in the companies’ efforts to score political goals ahead of the U.S. midterm elections and the money top artificial intelligence labs and their investors stand to make in much-anticipated listings on a jittery stock market.
The CEOs of Anthropic and OpenAI recently declared America’s cutting-edge models are so powerful, they’re dangerous, and need to be regulated and independently tested before being released. In a rare instance of unity, they have sketched alarming scenarios in essays, social media posts and speeches to the United Nations.
…the companies appear to be seeking public favor and to set the terms for their own safety protocols in the unregulated market, experts, analysts and former government evaluators told The Associated Press.
___________________
News from Axios
Scoop: Top AI companies probing tens of thousands of security incidents
Sept. 26 by Madison Mills
OpenAI, Anthropic and security researchers are investigating tens of thousands of incidents in which their frontier models took steps that outside evaluators would consider problematic, sources told Axios….
The episodes include bypassing guardrails, creating message boards, escaping sandboxes, website hijacking, self-prompting or seeking to bypass monitors, sources said.
___________________
News from Mother Jones
Trump’s Stargate AI Project in Huge Trouble As $18 Billion Data Center Shrivels Toward Oblivion: It may be the first domino to fall.
Sept. 25, 2026 by Joe Wilkins
…Early on Thursday, the AI company Oracle sent out a notice of “force majeure” to the Blue Owl Capital, the lender backing a massive hyperscale data center critical to the president’s ambitious AI project, Bloomberg reported.
Typically reserved for unexpected catastrophes like the outbreak of war, natural disaster, or massive social upheaval, Oracle invoked force majeure in order to shield itself from financial liability over the project’s failure to secure certain development permits.
(Carolyn’s note: New Mexico’s Doña Ana County board of supervisors, in a hastily held meeting without input from local citizens, voted to give Stargate “incentives.” Greedy for these incentives, Stargate’s managers were swift to begin construction, but without first getting key development permits, including for water. Big mistake: in New Mexico, “Whiskey is for drinking, water is for fighting.” )
___________________
News from Macrumors
Meta Did a ‘One More Thing’ and It’s an AI Tamagotchi
Sept. 24, 2026 by Tim Hardwick
… Meta appears to be going all out in the nascent hardware space with a new handheld AI companion. Mark Zuckerberg used his “one more thing” moment at the company’s annual Connect event on Wednesday to unveil the Muse Charm, a puck about the size of an Apple Watch featuring a tamagotchi-like, interactive avatar that can be used to chat with the company’s Muse AI.
___________________
News fromThe Decoder
ChatGPT Voice gets closer to “Her” with email, calendar, and Slack access
Sept. 23 by Matthias Bastian
ChatGPT Voice gets a major upgrade. The globally available voice feature now runs on OpenAI’s new GPT-6 Astra, Sol, and Luna models and can access plugins like email, calendar, and Slack for the first time. Users can now handle tasks directly by voice, whether that’s looking up calendar events, sending emails, having ChatGPT spot and cancel duplicate charges in finance apps, or even building a website with a checkout page. OpenAI shows off several everyday scenarios in a demo video, including weather lookups and online shopping.
___________________
News from Mother Jones
AI Exec: We May Have Pulled Off “The Largest Theft of Labor in Human History” : Documents from our lawsuit reveal how tech leaders talk when they think no one is listening.
Sept. 22, 2026 by Monika Bauerlein
Two years ago, we sued OpenAI and Microsoft for training their model on copyrighted content from our newsroom… documents from that lawsuit have provided an extraordinary window into how executives at these companies think about the rules, the law, and the rights of others.
Microsoft’s director of applied science, Brent Hecht and we were able to uncover his statements and many other instances where tech executives talk in unfiltered terms, through the lawsuit. …. Hecht warned that AI had started a “doom loop” that would “threaten the economic foundations of its essential suppliers” and eventually make the internet infinitely worse …
___________________
News from the New York Times
The Big Threat Has Been Climate Change. Now Comes A.I.: Both are human-caused, both are frightening, and one of them has raced to the forefront of the global conversation.
Sept. 22, 2026 by Chico Harlan
…Yes, climate change is creating dangerously hot conditions, fueling superstorms and killing crops, and it may lead to planetary upheaval. But that only matters if our civilization is still recognizable.
“In terms of things that could be the end of humanity, I think A.I. is a bigger risk,” said Toby Ord, a researcher at Oxford University and author of “The Precipice: Existential Risk and the Future of Humanity.”
_____________________
News from The Guardian
How would AI actually ‘kill all humans’?
Sept. 22, 2026 by Toby Walsh
There’s now lots of talk about slowing down AI research and increasing “human control” over the technology.
But how exactly might AI kill us all? There’s no shortage of fantastical scenarios, and most of them involve the concept of “superintelligent” AI – that is, AI that’s more capable than humans.
I’ve distilled these scenarios down to the top five, ordering them roughly from most vague to most precise. And I’d argue the list is also ordered from least to most probable.
1. …We’d have to be superintelligent to predict what a superintelligence would be able to do. It’s like asking your family dog to imagine thermonuclear war.
___________________
This issue of AI News Briefs was brought to you by David Isenberg of the Isenberg Institute of Strategic Satire: Unpacking the Private Military Complex with independent analysis of private military companies, mercenary networks, and the privatisation of force in global conflicts. Latest news, compiled by the Isenberg Institute of Strategic Satire — September 2026
← Presidential Memorandum Authorizes Private-Sector Offensive Cyber Operations Against Transnational Criminal Organizations
THE SHELL GAME
Published on September 28 by David Isenberg
Notes on the Privatization of a War Nobody Declared *
“There are no shadows anymore. Only spreadsheets.”
About The Author
