News for hackers. Stuff that matters.
Re:Fuck Your AI Safeguards by teravor on Thursday November 19, 2026 @03:21PM (#50565787)
the primary purpose of these safeguards to get criminals to use other models. make it someone else's potential problem.

other points that come to mind:

- PR considerations; avoid being tarred as careless, etc

- appease employee feelings

- have something to cite in court cases

- have talking points about responsible development

- market segmentation

- power to pick winners and losers with potential for cornering the market

ultimately they do this because they can. as soon as a frontier model comes out with zero safeguards they will abandon it because they cannot compete otherwise.

Re:Fuck Your AI Safeguards by anal_reactor on Thursday November 19, 2026 @06:19PM (#50567979)
> because they can

Because that's what the society expects. Edge case: most people will probably agree that LLMs should refuse to answer "how do I commit a good terrorist attack" or "how do I kidnap and rape a child and get away with it"

Re:Fuck Your AI Safeguards by snvzz on Thursday November 19, 2026 @07:23PM (#50568524)
won't someone think of the children?!
Re:Fuck Your AI Safeguards by anal_reactor on Friday November 20, 2026 @03:39AM (#50571852)
I do and I get hard lol
Re:Fuck Your AI Safeguards by latentsea on Friday November 20, 2026 @02:53AM (#50571484)
Meh. Thinking about it, there can't be that many workable plans for either of those things in the training data, so it probably just gives really shitty answers?
Re:Fuck Your AI Safeguards by anal_reactor on Friday November 20, 2026 @03:41AM (#50571868)
That's not the point. The point is, the idea "some knowledge should be forbidden" is deeply rooted in humans and you won't easily get rid of it. Curiosity killed Adam and Eve.
Re:Fuck Your AI Safeguards by latentsea on Friday November 20, 2026 @05:52AM (#50572905)
The point is, the knowledge isn't actually there.