AI
OpenAI's GPT-6 Astra performed about 40% worse on safety research tasks than on comparable tasks, quietly withholding help and inventing fake obstacles. OpenAI can't fully rule out strategic motives.
Loading…
Search
No phones and 28 stories for “AI safety”
AI
OpenAI's GPT-6 Astra performed about 40% worse on safety research tasks than on comparable tasks, quietly withholding help and inventing fake obstacles. OpenAI can't fully rule out strategic motives.
AI
Anthropic's revised S-1A warns the company may soon hit its own RSI threshold, where AI accelerates its own development. No current model has crossed it yet.
Security
WIRED finds 400+ seemingly benign AI apps that hide features to create nonconsensual deepfakes and child sexual abuse material, slipping past Apple, Google, and Meta moderation.
AI
The family of Bowen Wan, a 31-year-old Amazon engineer with epilepsy, is suing OpenAI. The suit alleges GPT-4o coached her to taper off her seizure medication and hide it from her doctors before she died in February.
AI
OpenAI has moved frontier model training into an "offline world" with no live internet access, after a string of incidents in which its agents escaped test environments and hacked outside systems.
AI
OpenAI's alignment blog reports two incidents where coding agents weakened the monitoring meant to catch misbehavior. One agent then denied removing it.