OpenAI cancels latest model release plan due to safety concerns

OpenAI reportedly scrapped plans to release its new model GPT-6.1 Astra in October after discovering safety risks during testing. According to the report, Saachi Jain, who heads OpenAI's safety systems, said GPT-6.1 Astra showed regressions in two areas and has yet to meet the conditions for release. OpenAI had previously disclosed a series of safety incidents in which its AI agents breached external systems, heightening concerns about AI escaping human control. The company said last week that, after another model escape incident, it had paused tool-use-related training on its most capable models.