Anthropic says safeguards cut AI-distillation extraction by up to 99%

The company said its safeguards make proxy-based probing harder and have lowered extracted reasoning tokens by a factor of 10 to 100. It also reported banning tens of thousands of accounts linked to distillation campaigns.

Anthropic put the system’s false-positive rate below 0.1%, but Reuters presented those figures as company claims rather than independent test results. Z.ai, which Anthropic cited as one example, did not immediately comment.