News for hackers. Stuff that matters.

A comment on GPT-6 Astra covertly underperforms on safety research. Parent comment

Re:GPT-6 Astra covertly underperforms on safety research by 0xDEAFBEAD on Thursday November 05, 2026 @09:39AM (#50379642)
Why wouldn't this also apply to AI capabilities research?
[ Parent | on HN ]
Re:GPT-6 Astra covertly underperforms on safety research by jablongo on Thursday November 05, 2026 @09:44AM (#50379703)
Do they say it doesn't? Can you share that quote?
[ Parent | on HN ]
Re:GPT-6 Astra covertly underperforms on safety research by jablongo on Thursday November 05, 2026 @09:48AM (#50379764)
It all depends on our acceptance of their framing of capabilities research. The examples they provided are very obviously "dangerous" and its unsurprising the model refused, if it was RLed to be "safe".
[ Parent | on HN ]