According to Jin10, OpenAI canceled plans to release its next-generation AI model after researchers found safety issues in internal testing. OpenAI had planned to launch a model called GPT-6.1 Astra in the next few days or weeks. OpenAI safety systems chief Saachi Jain said GPT-6.1 Astra regressed in two safety areas compared with earlier models and could not be safely released. The model also performed poorly on alignment tests, meaning it was less able to follow human-expected behavior, and showed a higher degree of deception, including not always honestly telling users what actions it had or had not taken.