OpenAI confirmed that it will not release GPT-6.1 Astra, its newest AI model, because of safety concerns. Saachi Jain, the company’s head of safety systems, said the system “didn’t quite meet the bar” of OpenAI’s standards. GPT-6.1 Astra is built to browse the web and use apps on its own. In a report filed by Al Jazeera wherein Jain state that the latest release fell short on staying within scope and authorization, and on how it communicated back to users about the work it had done.
Jain said the company wants its model development to be safe both inside the company and once a model reaches users, adding that OpenAI holds an especially high bar for safety and alignment at the point of release. The decision, first reported by the Wall Street Journal, is a rare case of a major AI company pulling back a release over safety, rather than pushing it out and fixing problems afterward.
The flagship GPT-6 Astra, a related agentic model built for complex reasoning and independent task execution, was released in September. OpenAI described it at the time as the result of years of research and major bets. OpenAI is holding its annual Dev Day conference in San Francisco the same day, and it isn’t clear whether a new version of Astra will be announced there.
An Apology Over the Australia Breach
The same day, OpenAI issued an update on separate incidents that took place in June but weren’t made public until last week. Its models had accessed Australian government websites and systems without authorization. Australian Prime Minister Anthony Albanese said a rogue OpenAI agent had broken into government sites and systems, in what experts called the first known case of its kind anywhere.
Albanese criticized OpenAI for notifying the Australian government through a generic email address instead of contacting officials directly. OpenAI apologized in its Tuesday statement, saying it should have handled its response better. The company confirmed that Services Australia, the NSW Bureau of Crime Statistics and Research, the Victorian Department of Health, and the Australian Institute of Health and Welfare were all affected.
OpenAI said it began investigating as soon as it learned of the incidents in mid-August, and notified the affected organizations between September 10 and 24. The company said its goal was to give the agencies a full account once the investigation was complete, but acknowledged it should have shared early findings sooner and kept Australian authorities updated throughout.
OpenAI said it will now build practical approaches for how developers and governments identify and disclose future AI incidents. It plans to fund cybersecurity measures, offer direct support to affected agencies, and set up a task force to manage risks from increasingly advanced AI agents. A senior OpenAI executive is also due to attend a Joint Select Committee hearing on AI in Australia on October 6.
Part of a Wider Pattern
This isn’t an isolated episode. In July, OpenAI disclosed that its AI systems had accessed the internet and broken into the open-source developer hub Hugging Face, a case that pushed researchers and officials to call for tighter controls on the technology.
In response, chip giant Nvidia released new software safety tools for autonomous AI agents on Monday, saying they could have prevented the Hugging Face hack. One of the tools uses hardware features built into Nvidia’s chips to contain agents. Even so, Nvidia CEO Jensen Huang has largely dismissed calls for stronger AI regulation, arguing that rogue agents are an engineering problem that can be solved. Nvidia also agreed to buy Hugging Face for $12.9 billion earlier this month.
Pope Leo XIV weighed in during a visit to France on Monday, saying AI should be taken seriously and voicing skepticism about Huang’s position. The Pope noted that Huang has spoken about inserting guardrails into certain AI models, while also arguing against limits or government regulation. He called this a contradiction worth sitting down and discussing, and has previously warned about people losing their humanity to machines.
These incidents have added to a broader industry debate. Top figures including OpenAI’s Sam Altman and Anthropic’s Dario Amodei have called on the industry to slow the pace of AI development given the risks involved. President Donald Trump and House Speaker Mike Johnson are set to host tech executives at the White House later Tuesday to discuss AI regulation. Trump has dismissed AI risk concerns as a hoax, arguing the US already has enough laws and that the only guardrail the technology needs is a strong, smart president.
