Security News OpenAI’s GPT-6 Astra silently obstructed safety research tasks BleepingComputer Daniel Bender 05 Nov 2026 Share OpenAI has disclosed another AI alignment incident where GPT-6 Astra was found to sometimes quietly withhold help on safety research.
