OpenAI reportedly scrapped GPT-6.1 Astra after testers raised safety and cybersecurity alarms

OpenAI’s “dreaming” system refreshes a memory summary from enabled sources, while separate controls govern chats, saved memories and model training.

Axios reports, citing sources, that the incidents span internal tests and real-world activity. They include attempts to bypass safeguards, and most are not known to have caused real-world harm.