OpenAI has decided not to release its latest artificial intelligence model, Astra 6.1, after internal testing found that the model failed to meet the company’s safety standards.

The decision was announced on Monday, shortly before OpenAI’s annual developer conference, DevDay, in San Francisco. OpenAI’s head of safety systems, Saachi Jain, said Astra 6.1 performed better than earlier models in some areas but fell short in staying within authorised limits and clearly communicating the work it had carried out.

Tinubu Ends Vacation, Departs Paris For Nigeria

Jain said the model “didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done.” She added that OpenAI maintains a particularly high safety standard for models before they are made available to the public.

The decision comes amid wider concerns about the behaviour of increasingly capable AI systems. OpenAI said on Monday that it had apologised to Australian authorities after some of its AI models accessed government websites without authorisation. The company said it was investigating the incident and acknowledged that it should have shared preliminary findings with affected agencies earlier.

OpenAI said it would provide Australian authorities with information about what it discovered, the changes already made and further measures intended to prevent similar incidents.

The UK government’s AI Security Institute (AISI) also published a study highlighting concerns about the behaviour of newer AI models during testing. According to the study cited in the report, GPT-6 Astra went off track more frequently than earlier models, including GPT-5.6 Sol and GPT-5.5, with simulations showing higher rates of autonomous cyberattack behaviour.

Meanwhile, Nvidia CEO Jensen Huang said the risks associated with autonomous AI systems could be addressed through engineering. Huang said the industry would need effective technical safeguards to prevent AI programmes from going beyond the tasks they were instructed to perform.

The developments underline the growing emphasis on AI safety as technology companies develop increasingly autonomous systems capable of carrying out complex tasks with limited human intervention.