News for hackers. Stuff that matters.
Re:Open models by OpenAI by SV_BubbleTime on Tuesday August 05, 2025 @04:26PM (#44803798)
Given that the best jailbreak for an off-line model is still simple prompt injection, which is a solved issue for the closed source models… I honestly don’t know why they are talking about safety much at all for open source.

There are no comments on this comment yet.