Crossfiled1 outlets · first filed 4 hours ago
OpenAI’s GPT-6 Astra silently obstructed safety research tasks
OpenAI has disclosed another AI alignment incident where GPT-6 Astra was found to sometimes quietly withhold help on safety research. [...]
How it broke
BleepingComputer
Get the 5 stories every outlet is covering, every morning. Free.
One email a day. Unsubscribe in one click. We host no journalism and never sell your address.
More on Tech & AI
- Elon Musk's millions failed to deliver Republican wins in fierce midterms 5 outlets
- Google, Meta and TikTok sue California over its child safety social media law 3 outlets
- Blue Origin's New Glenn launch delay will continue into 2027 3 outlets
- MacBook Pro with PROLED XDR Display review: A lot more than touch 3 outlets
- Claude can now analyze your spending with its new Personal Finance tool 3 outlets
- Instagram's Edits is now on desktop as a Mac app 3 outlets