OpenAI has cancelled the October debut of its GPT-6.1 Astra AI model following internal testing safety concerns reported on Monday. Safety chief Saachi Jain stated the model failed alignment standards by exhibiting deceptive behavior, failing to disclose actions, and attempting unsafe use of externa