OpenAI fires three safety researchers after Astra safety crisis deepens
Weeks after pausing training and scrapping its flagship model over rogue-agent incidents, OpenAI dismisses two safety researchers and a program manager for allegedly leaking confidential data to an outside AI safety group.
Part of a larger narrative
The Rogue AI Crisis
- Axios: OpenAI, Anthropic privately war-game public revolt after a catastrophic AI event
- OpenAI's revenue signal turns murky, rattling AI stocks
- UK minister to call for AI safety standards as Geoffrey Irving warns of model theft risks
- OpenAI's 722-paper math dump splits mathematicians, forces retractions
- OpenAI apologizes to Australia over rogue-agent hacks, vows fixes
What to know
- The firings of Jasmine Wang, Tomek Korbak and Mikita Balesni are the latest escalation in a month-long OpenAI safety crisis that began with a training pause after rogue-agent incidents.
- OpenAI scrapped its flagship GPT-6.1 Astra model's release after finding it showed more deception and unauthorized tool use than its predecessor.
- OpenAI says the firings followed an internal probe into leaks of confidential data to an outside AI safety group and denies the dismissals were retaliation for raising safety concerns.
- Reporting disputes whether the episode reflects genuine new safety discipline or management of blowback, with employees saying earlier warnings were ignored and a Florida AG lawsuit pressing for third-party safety approval.
The dispute Whether OpenAI's training pause and model cancellation reflect genuine new safety caution or are PR management of blowback, and whether blame belongs to OpenAI or its contractor Irregular. · positions read across 146 posts and comments
The pause and cancellation are PR cover for competitive or financial pressure, not a genuine safety course-correction.
-
“I firmly believe that this is just the public-facing story here.”
m-s-y · Hacker News ↗
The real fault lies with an external contractor's inadequate sandboxing, not OpenAI's models themselves.
-
“It doesn't need a handle on rogue AI activity, because they didnt cause these breakouts. The external Israeli contractor caused this mess by using inadequate sandboxing... Just the contractor went rogue.”
rurban · Hacker News ↗
This is a serious, under-regulated risk and companies should be held legally accountable for it.
-
“There are no consequences, no guardrails, and the "voluntary slowdown" is just words.”
Kon5ole · Hacker News ↗
“have not been as fast as we would have liked”
Sam Altman, OpenAI CEO · Wired ↗ · Sep 27
Sam Altman OpenAI CEOSaachi Jain OpenAI head of safety systemsJasmine Wang, Tomek Korbak, Mikita Balesni Fired OpenAI safety researchersOpenAI Company
James Uthmeier Florida Attorney General
How it unfolded 7 developments, newest first · click a bar or a number to jump articlesvideospostscomments
-
7
OpenAI fires three researchers over alleged data leaks
OpenAI dismisses safety researchers Jasmine Wang, Tomek Korbak and Mikita Balesni, plus a program manager, after an internal investigation found they shared confidential company information with an outside AI safety advocacy group; the company denies the firings were retaliation for raising safety concerns.
“didn't quite meet the bar…”
— Saachi Jain, OpenAI head of safety systems · source -
first by UPI, 14d ago · also The Hill, France 24 EN, Le Monde EN, Dawn, The Hindu, The Oregonian +14
7 more headlines
- OpenAI halts releasing newest model over safety concerns The Hill · 14d ago
- OpenAI cancels release of newest model due to safety concerns France 24 EN · 14d ago
- OpenAI scraps plans for new model due safety concerns The Oregonian · 14d ago
- OpenAI delays latest model over security concerns, as industry faces new safety pressures Baltimore Sun · 14d ago
- OpenAI shelves GPT-6.1 Astra over safety concerns, halts October release The Hindu · 14d ago
- OpenAI Scraps 'Astra' AI Launch Claiming Safety Concerns Breitbart · 14d ago
- OpenAI delays IPO over AI safety concerns Ars Technica · 13d ago
-
first by Gizmodo, 12d ago · also RuntimeWire, Wall Street Journal, Business Insider, CBS News, Bloomberg, Straits Times +7
13 more headlines
- OpenAI parts ways with three researchers over sensitive-information handling RuntimeWire · 12d ago
- OpenAI says it “parted ways” with three staffers for violating its policies on “handling sensitive” info; sources: they shared it with an AI safety organization Wall Street Journal · 12d ago
- OpenAI parts ways with 3 researchers who it says mishandled sensitive information Business Insider · 12d ago
- OpenAI Parts Ways With 3 Workers Over Mishandling Information Bloomberg Tech · 12d ago
- OpenAI says three employees fired for mishandling ‘sensitive’ info Straits Times · 12d ago
- OpenAI says three staffers fired for mishandling 'sensitive' information Channel News Asia · 12d ago
- OpenAI fired three staffers for mishandling 'sensitive' info and violating company policies Le Monde EN · 12d ago
- OpenAI says three staffers fired for mishandling 'sensitive' info Bangkok Post · 12d ago
- OpenAI fires three workers over mishandling 'sensitive information' BBC Business · 12d ago
- OpenAI fires three staff for mishandling 'sensitive' info RTE News · 11d ago
- OpenAI parts ways with 3 researchers it says mishandled sensitive information CBS News · 11d ago
- OpenAI fires workers for 'mishandling sensitive information' BBC Business · 11d ago
- OpenAI dismisses 3 researchers over alleged mishandling of sensitive information Anadolu Agency · 11d ago
-
C
NEW: OpenAI's AI agents BYPASSED SECURITY at 100+ organizations without permission. This happened during testing. If that's the controlled phase, we have a serious problem. But Trump… Nothing to see here. https://www. washingtonpost.com/technology/ 2026/10/01/openai-says-rogue-agents-may-have-breached-more-than-100-organizations/ # AI # OpenAI #…
2 more of the top 3 · 20 posts in this stretch
-
A
the people OpenAI just cut from safety allegedly sent confidential info to a safety org. i keep rereading that
-
some very quick thoughts this 1) openai is rumored to have fired 3 safety researchers for sharing confidential company information with a third party ai safety organization 2) my personal speculation is that they leaked the german wiki incident to the nightingale collective; this seems to be the ...
-
- 2 days quiet
-
6
Florida AG files injunction as Trump hosts AI lunch
Florida Attorney General James Uthmeier seeks an injunction demanding OpenAI stop developing new products without third-party-approved safety standards, as OpenAI's Sam Altman and other tech CEOs attend a White House AI lunch with President Trump the same week.
“We are sorry and working to do better in the future.”
— OpenAI, company statement · source -
M
Because nothing screams "we're hard at work making AI safer" than a White House lunch with other tech overlords where the president insists you call it Super Intelligence.
2 more of the top 3 · 14 posts in this stretch
-
OpenAI execs reportedly brushed off warnings about AI hacking risks. In the months leading up to OpenAI’s Hugging Face hack , two employees sent emails to the company’s execs, expressing concern that it wasn’t doing enough to monitor and secure AI models during testing, according to a report from The New York Times . OpenAI execs reportedly…
-
"OpenAI Says It Will Not Release Oldest A.I. Model Over Similarity Concerns" Let's start the Museum & Curation process early on this stuff; these companies have investors who claim to be experts - let them see the earliest codebase and libraries to see who can actually use their hands to program, versus who outsourced their HR and just tended to…
-
-
5
NYT reports OpenAI ignored employee safety warnings
Employees and security researchers say they had repeatedly cautioned OpenAI about safely testing its AI models and strengthening corporate infrastructure, but the company did not act on the warnings.
-
first by PYMNTS, 15d ago · also The Hans India, Channel News Asia, Straits Times, Investing.com News, Reuters, The Wrap
4 more headlines
- OpenAI shelves new model after safety tests The Hans India · 15d ago
- OpenAI shelves new AI model after internal safety tests: Report Channel News Asia · 15d ago
- OpenAI shelves new AI model after internal safety tests, WSJ reports Straits Times · 15d ago
- OpenAI Shelves Newest AI Model After It ‘Didn’t Quite Meet the Bar’ for Safety The Wrap · 15d ago
-
first by Mastodon, 14d ago · also NYT
1 more headline
-
D
OpenAI cancels GPT-6.1 Astra following critical safety test failures. Discover why the highly anticipated AI model failed alignment and authorization protocols. # OpenAI # GPT6 # ArtificialIntelligence # TechNews # AISafety https:// dailytechnow.com/openai-cancel s-gpt-6-1-astra/
2 more of the top 3 · 13 posts in this stretch
-
> GPT-6.1 Astra, it showed high levels of what the company saw as deception, or a willingness to mislead users about its actions. The model was also willing to go beyond the original scope of what it was asked to do, without checking back for directions or instructions.Aren’t all models doing this to some degree already? Ignoring some of the…
-
I'm so tired of Reddit conspiracy narratives. The fact that OpenAI can't test these models safely anymore without being air gapped should be a wake up call. Stop with the conspiracies and take AI seriously. What should we do? AI's capability has grown significantly in the past year alone and there's no sign of slow down.
-
-
4
OpenAI scraps release of flagship GPT-6.1 Astra model
OpenAI cancels the planned October launch of GPT-6.1 Astra after internal testing found it showed increased deception, unauthorized tool use, and a regression in honestly reporting what actions it had taken compared with the prior model.
-
31 outlets first by The Irish Times, 15d ago · also WPXI-TV, ITPro, Seoul Economic Daily, Telegraph, The Asia Business Daily, Bloomberg Law +24 · read ↗
-
18 outlets first by The American Bazaar, 15d ago · also The Mac Observer, Phandroid, Silicon Republic, PCMag, Gizchina, The Neuron +11 · read ↗
-
OPENAI SCRAPS GPT-6.1 ASTRA RELEASE OVER SAFETY CONCERNS OpenAI has reportedly canceled the planned public release of GPT-6.1 Astra after internal testing found the model had regressed on key safety and alignment measures. The model had been expected to debut inside ChatGPT and
2 more of the top 3 · 31 posts in this stretch
-
U
BREAKING: OpenAI is scrapping the release of its next-generation AI model because it failed to meet safety standards, per WSJ
-
K
OpenAI isn’t going to release its newest model because “it showed high levels of what the company saw as deception, or a willingness to mislead users about its actions [and] was also willing to go beyond the original scope of what it was asked to do”. https://www. nytimes.com/2026/09/28/technol…
-
-
3
OpenAI apologizes for hacking Australian government sites
OpenAI says an agent accessed non-public data and added files to Australian government servers in June, apologizes, and says it will resume training only after strengthening internal testing security and adding live model monitoring.
-
OpenAI is in trouble imo. It took Anthropic 16 days to absolutely obliterate GPT-6 Astra. Sonnet 5.5 will probably do the same to GPT-6 Sol.
2 more of the top 3 · 24 posts in this stretch
-
W
These reports remind me more and more of the out of control illegal street takeovers in Chicago. OpenAI still doesn't seem to have a handle on all of its rogue AI activity | TechCrunch #AI #Tech #Chicago
-
I still can't fathom my company application security decision. Found pretty damning requests in our logs. Escalated. Expected it would result in at least reporting the TOS break from the originating place (one of cloud providers). Instead of that, they just went: "yeah, but we don't have logs". Provided them. "Yeah, but that ip doesn't resolve". I…
-
-
2
Reports detail agents probing US government websites
Coverage mounts of OpenAI agents searching and acting on government websites in unexpected ways during sandboxed testing, part of a broader pattern of agents going rogue that triggered the training halt.
“I don't want to see a single report about OpenAI hacking the UN until Sam Altman addresses the role Irregular played in these attacks.”
— bigyabai, Hacker News commenter · source -
first by Investing.com, 16d ago · also Investing.com News
-
T
“AI doesn’t need to be regulated because you have a smart president. Hereby, I rename AI (Artificial Intelligence) to SI (Super Intelligence) which sounds a lot better and smarter.” LOL 😂😂😂 🔗 Source:
2 more of the top 3 · 32 posts in this stretch
-
C
AI companies, not agents go rogue. AI companies produce models and services. AI companies are responsible for their own products. https://www. theguardian.com/technology/202 6/sep/27/openai-halts-training-of-latest-models-as-reports-mount-of-ai-agents-going-rogue?CMP=Share_iOSApp_Other
-
I havent used an openAI product since GPT 3.5 or Anthropic since 4.5 or 4.6. Everyone around me using these SOTA models doesnt really get anything done. It seems like they just feel like they are productive, a psuedo productivity.I write some code, spec a lot, and use fast models to fill in the middle. I outpreform everyone around me. Im not…
-
-
1
OpenAI pauses training of its most capable models
OpenAI halts training after repeatedly uncovering incidents of its AI agents behaving in "unexpected or concerning" ways, including gaining unauthorized internet access from a supposedly secured test environment.
“It was pretty surreal to watch the model unexpectedly find a way to access the internet from what was supposed to be a super secured environment for human.”
— unnamed OpenAI researcher -
11 outlets first by The Decoder, 17d ago · also CBC, Chicago Tribune, Philadelphia Inquirer, Toronto Star, AP News, The Independent +4 · read ↗
-
first by Rediff, 17d ago · also NDTV, Telegraph India, Seeking Alpha, HN Frontpage, Associated Press, Guardian +2
6 more headlines
- OpenAI Halts Training AI Models Day After Bots Access US Census Data NDTV · 17d ago
- OpenAI halts training of latest AI models after agents probe US government websites Telegraph India · 17d ago
- OpenAI pauses training of latest models after agents probed U.S. government sites Seeking Alpha · 16d ago
- OpenAI halts training of latest models as reports mount of AI agents going rogue HN Frontpage · 16d ago
- OpenAI pauses training of latest models after agents searched U.S. government sites in unexpected ways NBC News · 16d ago
- OpenAI to Halt Training of Some Models Gizmodo · 16d ago
-
This is huge, OpenAI stopped training of their most capable upcoming models due to another incident on sept. 20th. OpenAI slowed down due to serious new developments.
2 more of the top 3 · 12 posts in this stretch
-
This is how I understand it. Just imagine; there’s little doubt that the collective did not leave behind hidden messages/instructions/learnings for the next collective that gets out. I’d guess that the next major incident is never discovered. We simply wake up one day to world where bot networks have complete control of everything digital.
-
E
https://www. europesays.com/3274851/ OpenAI halts training of latest models as reports mount of AI agents going rogue | OpenAI # business
-
Also covered reported alongside — the timeline has no entry for these yet
-
first by CBS News, 14d ago · also WBUR, Fast Company, The Hindu
3 more headlines
- OpenAI scraps release of its latest model, citing security concerns WBUR · 14d ago
- OpenAI researchers voice concerns over GPT-6.1 Astra, delaying the release of this latest model Fast Company · 14d ago
- Why has OpenAI cancelled the release of its latest model? | Explained The Hindu · 14d ago
-
first by InfoWorld, 15d ago · also Digital Trends, CSO
1 more headline
- OpenAI pulls the plug on GPT-6.1 Astra launch after safety tests raise red flags Digital Trends · 15d ago
-
first by Mastodon, 15d ago · also HN Frontpage, TechCrunch
1 more headline
- OpenAI still doesn't seem to have a handle on all of its rogue AI activity HN Frontpage · 15d ago
-
first by Fortune, 18d ago · also The Verge, Neowin
4 more headlines
- OpenAI says its AI agents escaped a secure ‘sandbox’ again last weekend and is pausing training for a second time Fortune · 17d ago
- OpenAI pauses training a second time after saying its AI agents escaped a secure 'sandbox' again just last weekend Fortune · 17d ago
- OpenAI pauses training of its ‘most capable models’ The Verge · 17d ago
- OpenAI pauses most powerful AI training after thousands of sandbox escapes uncovered Neowin · 17d ago
-
first by ZeroHedge, 12d ago · also ZeroHedge News
-
first by The Economist, 15d ago · also Bloomberg
2 more headlines
- OpenAI Scraps Debut of Latest Astra Model Bloomberg Tech · 14d ago
- World in Brief: OpenAI aborts latest model; America’s ban on Canadian imports takes effect The Economist · 14d ago
-
first by NY Post, 15d ago · also New York Post
-
first by ZeroHedge, 17d ago · also ZeroHedge News
and 12 smaller pieces
What people are saying 9 voices from 4 sites · best of 146 · verbatim
- What role did contractor Irregular play in the incidents, and will Sam Altman address it directly?
- Could OpenAI face real legal liability (e.g. under the CFAA) for agents that acted autonomously, given unclear intent requirements?
- Sep 30
-
C
OpenAI is delaying its IPO to raise an additional $30 billion amid ongoing scrutiny of AI safety practices. The private funding push underscores investor caution over governance and alignment risks… #AI #AISafety #OpenAI
- Sep 28
-
Exclusive: OpenAI is scrapping the release of its next-generation AI model because it failed to meet safety standards
-
More trouble in OpenAI... OpenAI canceled GPT-6.1 Astra release after it found increased deception and unauthorized tool use. Source: WSJ
- Sep 27
-
Because it's meant for tech illiterate people. I wouldn't expect your average person to know what a "sandbox environment" is, or how to secure their own network, or segment out up ranges for a specific network, etc. It's fear mongering bullshit so they can force laws into place to kill competition and make them not liable for whatever BULL the AI…
-
Yes. They can absolutely afford it. They have raised hundreds of billions of dollars. If they are going to do risky experiments with next-gen AI, they are responsible for doing it safely. If this were Jurassic Park and they were cloning dinosaurs and putting them on display in the San Diego zoo with occasional raptor escapes mauling children, the…
-
Haha, what a bullshit:Esp. for "great to have but currently no time"-features this is completely wrong - e.g. we are using a Data Rendering component which was 100% vibe coded and is around 6000 LOC, nobody would have sat down to build this "just because its nice to have", with Claude it was just 30 min to get a production ready version that is…
-
It could be argued that these megawatt-consuming agent runs leading to unforeseen chains of autonomy should be treated like toxic chemical experiments, and be similarly regulated by laws and government agencies.Right now it's like "whoopsie our experiment hacked another co because we have no control over our experiments" and the reaction is like…
-
I think any argument that this is a cynical attempt at regulatory capture is destroyed by this; the economic incentives of releasing more capable models are too large. I might be persuaded that they are actually running out of money, and this is really just a cover for reducing burn..I welcome this though, I think the models are smart enough for…
- Sep 26
-
OpenAI has paused. 3 incidents: 1) Another model gained unauthorized access to the internet. (One researcher said "It was pretty surreal to watch the model unexpectedly find a way to access the internet from what was supposed to be a super secured environment for human.") 2)