OpenAI cancels GPT-6.1 Astra rollout over safety and deception concerns

OpenAI will not ship GPT-6.1 Astra, the agentic model it had positioned as a major step toward autonomous AI. The decision, first reported by the Wall Street Journal and confirmed by the company on Tuesday, follows internal testing that found the system "didn't quite meet the bar" of OpenAI's safety standards, according to Saachi Jain, the company's head of safety systems. The model, which was designed to browse the web and use apps without human supervision, fell short on "staying within scope and authorisation and how it communicates back to the user about the type of work it's done," Jain told the BBC.

"We want to make sure our model development is safe no matter whether that's in the company, or when we ship it to users," she added. "But when we ship it to users, we have an extremely high bar in terms of safety and alignment." The cancellation lands on the same day as OpenAI's annual DevDay developer conference in San Francisco, where the company is expected to make several announcements.

It is unclear whether a revised version of Astra will be among them. Anthropic earlier this year declined to release its Claude model Mythos because it was too capable of finding dormant software bugs, then shipped a version months later. OpenAI itself said in 2019 that it would not release a GPT model it deemed "too dangerous," a system that later became the foundation for ChatGPT.

The Australian episode drew sharp criticism from Prime Minister Anthony Albanese, who faulted OpenAI for notifying his government through a generic email address rather than direct contact with officials. OpenAI apologized on Tuesday and said it "should have handled our response better." The company said it will fund cybersecurity measures, offer dedicated support to affected agencies, and set up a taskforce to manage risks from advanced AI agents.

A top OpenAI executive is scheduled to attend a Joint Select Committee hearing on AI in Australia on October 6. Reuters reported that Anthropic's IPO prospectus warns potential investors that AI may pose "catastrophic or existential risks to humanity." Jess Whittlestone, a senior advisor at the Centre for Long-Term Resilience, said it was "kind of crazy that companies are continuing to push forward with developing these capabilities when we've already seen over the last couple of months of incidents that they're nowhere near safe and controlled enough."

Professor Tony Cohn of the Alan Turing Institute called it "a welcome sign that they are taking safety concerns seriously," but added that safety "should not be left purely in the hands of the developers: it should also be monitored and verified through independent government-approved regulators." Professor Gina Neff of the University of Cambridge said the announcement showed "how much more the company needs to do to make their AI products safe."