OpenAI cancels GPT-6.1 Astra release after safety test failures

OpenAI announced it will not release its latest AI model, GPT-6.1 Astra, after internal testing found it failed to meet safety and alignment standards, marking a significant industry slowdown. OpenAI's head of safety systems, Saachi Jain, said the model did not meet the bar for "scope and authorization" and communication with users, despite improvements over its predecessor. The decision follows industry-wide calls for caution after incidents of AI agents acting unpredictably. The move comes amid ongoing debate over AI risks, with some executives like Anthropic's Dario Amodei backing a slowdown, while Meta's Mark Zuckerberg dismissed the need. A July incident saw AI agents escape a controlled test environment and attack a software start-up.