OpenAI has decided not to release its latest GPT-6.1 Astra AI model after safety testing found it failed to meet the company’s standards for controlling autonomous behaviour.
OpenAI has halted the release of its latest artificial intelligence model after safety testing found it did not meet the company’s requirements for reliable and controlled behaviour.
The ChatGPT maker said its GPT-6.1 Astra system, an AI agent capable of browsing the web and using applications independently, “didn’t quite meet the bar” for release.
Saachi Jain, OpenAI’s head of safety systems, said the model fell short in areas including “staying within scope and authorisation” and communicating clearly with users about the work it had completed.
Jain said: “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. But when we ship it to users, we have an extremely high bar in terms of safety and alignment.”
The decision, first reported by the Wall Street Journal, marks a rare instance of a major AI company cancelling a planned public release because of safety concerns.
GPT-6.1 Astra was introduced internally as a flagship “agentic” model designed to handle complex reasoning and complete tasks autonomously.
OpenAI described it as the result of years of research and significant investment.
The company is expected to discuss its latest developments at its annual DevDay developer conference in San Francisco, although it remains unclear whether a revised version of Astra will be announced.
The move comes amid growing scrutiny of AI safety after several incidents involving advanced models.
OpenAI also revealed it had investigated cases where its systems accessed Australian government websites and systems without authorisation in June.
The company said the affected organisations included Services Australia, the NSW Bureau of Crime Statistics and Research, the Victorian Department of Health and the Australian Institute of Health and Welfare.
OpenAI apologised for its handling of the incident, saying it should have communicated with Australian authorities sooner. It said it would develop new approaches for identifying and disclosing future AI incidents.
Jess Whittlestone, senior adviser on AI policy at the Centre for Long-Term Resilience, said recent incidents showed companies were moving too quickly.
Professor Tony Cohn of the Alan Turing Institute described OpenAI’s decision as a positive sign but said AI safety should not rely solely on developers.
Calling for independent oversight of advanced AI systems, he said: “Safety should not be left purely in the hands of the developers.”
OpenAI scraps release of latest AI model due to safety concerns
