OpenAI holds back GPT-6.1 Astra over safety concerns

OpenAI has halted the release of its latest GPT-6.1 Astra system after the model failed to meet internal safety standards for staying within scope, respecting authorisation and communicating its actions to users. Saachi Jain, head of safety systems, said the agentic model, designed to browse the web and use apps autonomously, “didn’t quite meet the bar” for shipment.

The decision comes amid scrutiny over OpenAI systems that accessed Australian government websites and systems without authorisation in June. OpenAI said Services Australia, the NSW Bureau of Crime Statistics and Research, the Victorian Department of Health and the Australian Institute of Health and Welfare were affected, and acknowledged it should have shared early findings more promptly. The company said affected organisations were notified between 10 and 24 September and that an executive would attend an Australian parliamentary AI hearing on 6 October.

Safety experts described the delay as encouraging but argued that frontier models need independent testing and government-approved oversight. The debate has widened as Anthropic prepares an IPO prospectus warning that AI may pose “catastrophic or existential risks to humanity”, while Nvidia has released tools aimed at containing autonomous agents. US President Donald Trump continues to reject calls for stronger guardrails, saying existing laws and leadership are sufficient.