OpenAI has cancelled the planned public release of its latest artificial intelligence model, Astra 6.1, after internal testing found that the system did not meet the company’s safety standards.
The model showed improvements in some areas but reportedly performed poorly in tests measuring whether it stayed within authorised limits and accurately communicated the work it had carried out. OpenAI’s head of safety systems, Saachi Jain, said the company maintains a particularly high standard for safety and alignment before releasing models to users.
The decision comes amid growing concerns over the behaviour of increasingly autonomous AI systems. Recent testing and incidents have raised questions about whether advanced models can reliably follow restrictions when given access to external tools and websites.
OpenAI had been expected to discuss new developments at its annual developer conference in San Francisco, but it remains unclear whether a revised version of Astra 6.1 will feature in the company’s announcements.
The cancellation highlights the growing importance of safety testing as technology companies compete to develop increasingly capable AI systems while attempting to maintain stronger controls over how those systems operate
Leave a comment