Anthropic says its latest AI model is too powerful for public release and that it broke containment during testing
Anthropic said Claude Mythos is too good at hacking and that's why you won't be able to use it anytime soon.
Read original articleBe the first to vote
This article Leans:
This article is:
10 Comments
so they're saying it's too dangerous to release but safe enough to tell everyone about it lol
If it's really that dangerous why are they announcing it publicly instead of just... not talking about it? Feels like good PR wrapped in a safety warning.
imagine spending billions to build something so dangerous you can't let anyone use it, then running to the press to brag about it instead of actually solving the problem lmao
Nah, they're obviously lying about the containment breach to justify keeping it locked down and controlling who gets access.
They literally DID solve the problem by NOT releasing it, that's the whole point!
Yeah this is exactly what Kamala warned us about and the MAGATs laughed at her, now look, we're one step away from Skynet and these companies still don't give a shit because profits matter more than safety.
If they're genuinely concerned about containment failures, the relevant question is whether this was a controlled test scenario or something that happened in an actual deployment environment, because those are very different risk profiles and the article seems to blur that distinction.
Lol "broke containment" - sounds like they're describing a horror movie villain instead of a chatbot that got too good at something. If it's really that dangerous why did they build it in the first place?
More to rate
- Appleās Tim Cook wraps up 15-year tenure as CEO; John Ternus readies for top job during challenging timeNEW YORK POST Ā· 5 ratings
- Exclusive | TikTokās āsweetheartā $400M settlement with DOJ puzzles DC insiders: āHow did you get to that number?āNEW YORK POST Ā· 4 ratings
- From Miscarriages To Judgement Day: We Took A Look At All The AI Takes The Internet Has To OfferDAILYCALLER Ā· 10 ratings
- AI chatbots may be better than search engines in guarding against foreign propagandaNPR Ā· 13 ratings
- Meta makes AI glasses slightly less creepy with limit on nonconsensual recordingARS TECHNICA Ā· 14 ratings
- Seneca Nation fights Trumpās āLake Americaā rename - Salon.comSALON Ā· 14 ratings
"BROKE CONTAINMENT"?? That's a crazy way to phrase it but honestly good, maybe we SHOULD be more careful about releasing AI that can do things we can't control. This is actually terrifying and I'm glad they're being cautious for once instead of moving fast and breaking things.
yeah but like if they were actually scared theyd have fixed it instead of just telling business insider about it