Anthropic says distillation attacks fell 90–99% after banning tens of thousands of accounts
Anthropic says it has cut attempts to steal Claude's reasoning data by 90–99%, banning tens of thousands of accounts in the process, and calls the fight far from over.
Anthropic says it has cut attempts to steal Claude’s reasoning data by 90 99% The crazy part is that some labs were running thousands of requests just to extract Claude’s thinking and use it to train their own models Anthropic has now strengthened its defenses, banned tens of thousands of accounts, and says the fight is far from over Honestly, distillation itself isn’t the problem. It’s how AI improves and becomes cheaper The problem is copying frontier capabilities without the same investment in safety This is going to be a long cat and mouse game.
This post quotes another post: x.com/anthropicai/status/2109795898909950247