New safeguards sharply reduce AI distillation attacks, says Anthropic

Anthropic said ​it had made it harder to use proxy networks to repeatedly probe Claude's protected reasoning, blocking extraction attempts and hindering large-scale ‌distillation efforts

Anthropic said on Monday new measures designed to prevent rivals ​from copying the capabilities of its Claude AI models have reduced data harvested in so-called distillation attacks by about 90% to 99%.

 

Model distillation has emerged as a key concern for leading ‌AI companies due to ​intensifying US-China competition, with ​firms including Anthropic, OpenAI and Google arguing the practice enables ​rivals to reproduce advanced capabilities more quickly and at lower cost.

 

Anthropic said safeguards restricting access to Claude's reasoning have sharply reduced extraction attempts, with "the number of ​reasoning tokens extracted by distillers" now "10 to 100 times ‌lower than it was before."

 

 

The California-headquartered company said ​it had made it harder to use proxy networks to repeatedly probe Claude's protected reasoning, blocking extraction attempts and hindering large-scale ‌distillation efforts.

 

Anthropic said ​its safety system is "roughly ‌as robust as those that prevent Claude from helping ‌build bioweapons" and has "a false positive rate of ​less than 0.1%."

 

Anthropic banned tens of thousands of accounts linked to distillation campaigns and ​cited Chinese AI startup Z.ai , also known as Zhipu, as an example of a ‌company that attempted to distill capabilities from Claude while ‌developing its GLM-5.3 model, warning such practices pose safety and national security risks.

 

Z.AI did not immediately respond to a request for comment.