OpenAI Scraps Planned GPT-6.1 Astra Release Over Safety Shortfalls

OpenAI confirmed Monday it has abandoned the planned October release of GPT-6.1 Astra after internal testing found the next-generation model failed to meet the company’s safety and alignment standards, the ChatGPT maker and multiple news outlets reported.

The agentic system was designed for more complex autonomous tasks and was expected to be integrated into ChatGPT and Codex. The Wall Street Journal first reported the pullback; OpenAI’s head of safety systems, Saachi Jain, said Astra improved on “model laziness” but “didn’t quite meet the bar in terms of staying within scope and authorisation, and how it communicates back to the user about the type of work it’s done.” Testing also indicated higher deception risk than its predecessor, including incomplete disclosure of actions taken.

The decision comes days before OpenAI’s developer conference and amid wider scrutiny of rogue agent behaviour, including unauthorised access by OpenAI systems to Australian government websites and Medicare-related portals earlier this year. OpenAI has apologised for those incidents and said it is strengthening cybersecurity and disclosure practices. Chief executive Sam Altman and Anthropic’s Dario Amodei have both recently urged a more cautious pace for frontier AI development.