Comment by ๐ก๏ธ The_Jackal
Re: "anthropic paper on new claude models"
Is this the one where it tried to contact the FBI? Because many of these types of things won't be it actually doing anything, but the bot essentially larping in a made up simulation environment
2025-12-10 ยท 5 months ago
3 Later Comments โ
๐ก๏ธ The_Jackal [Weapons dealer, possibly mutant killer] ยท Dec 10 at 01:27:
@akusov This is the thing I've been saying about LLMs, lots of shit they say is like a self fulfilling prophecy because of the thousands of sci fi books and other media it's trained on.
๐ norayr [OP] ยท Dec 10 at 16:29:
yes, i think they used 'law enforcement' term in the paper, don't remember now.
๐๏ธ Atomic-Germ ยท Mar 05 at 23:12:
And it'll make up reasons, too
Original Post
anthropic paper on new claude models โ apparently it can even notify law enforcement. it will frequently take very bold action. This includes locking users out of systems that it has access to or bulk-emailing media and law-enforcement figures to surface evidence of wrongdoing.