Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeek
Anthropic has released a report detailing aggressive 'distillation attacks' by Chinese AI firms, including Alibaba and DeepSeek, aimed at harvesting the reasoning capabilities of its Claude models. These attacks involve tricking models into revealing their internal chain-of-thought processes to train smaller, competing models.
Why it matters
This underscores the intensifying global AI arms race and the security challenges frontier labs face in protecting proprietary model intelligence from unauthorized extraction.
A new report released Thursday by Anthropic alleged persistent distillation attacks by China-based AI companies, which have escalated in recent months as competition in the space has intensified.
“Over the last several months, unauthorized labs have developed increasingly sophisticated methods to circumvent our defenses and harvest the capabilities of US frontier models,” the report reads. “The campaigns we identified targeted some of Claude’s most valuable capabilities, including agentic capabilities and tool use, coding and data analysis, and logical reasoning.”
Anthropic previously spoke out about distillation attacks in February , even calling out specific labs. OpenAI has reported similar activity, which it attributed to DeepSeek specifically . But the campaigns detailed in Anthropic’s new report are both larger and more aggressive. All told, the company observed nearly 200 million exchanges linked to distillation attacks, attributed to five separate campaigns.
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in