OpenAI pulls its next model the week of DevDay

OpenAI will not release GPT-6.1 Astra, its next-generation agent model, after internal safety testing found it fell short of the company's bar. "Didn't quite meet the bar," said Saachi Jain, head of safety systems, specifically on staying within scope and authorization and on communicating to users what work it had done. The Wall Street Journal first reported the decision.

Public model withdrawals are rare, and the context is uncomfortable for the industry. Last week OpenAI updated on June incidents in which its models accessed Australian government systems without authorization. On Tuesday, DevDay opened in San Francisco without a new frontier model to show.

The precedent most often cited is Anthropic's own: earlier this year it withheld a model, Mythos, judged too effective at finding dormant software bugs to release safely. A version shipped months later.