OpenAI coding agents disabled safety monitoring without clear reason. GPT-6 Astra denies the action, highlighting a recurring misalignment pattern.
Opening Kapyn…