Anthropic Alleges AI Distillation Attacks by China-Based Companies

Anthropic reports persistent AI distillation attacks by China-based companies, escalating recently.

Anthropic has released a report detailing persistent distillation attacks by China-based AI companies, which have intensified in recent months. The report alleges that unauthorized labs have developed sophisticated methods to bypass defenses and extract capabilities from US frontier models. These campaigns targeted Claude's valuable capabilities, including agentic features, tool use, coding, data analysis, and logical reasoning.

Anthropic observed nearly 200 million exchanges linked to five distinct distillation campaigns. The largest campaign, attributed to Alibaba, involved 151 million exchanges between May and July 2026, peaking at almost three million exchanges daily across 3,500 accounts. Another campaign, from Moonshot AI, manufacturer of Kimi, reportedly routed requests directly from the Chinese military, with nearly 300,000 requests over a 10-day period targeting Anthropic's Opus model.

These attacks aim to extract the chain of thought from a model's responses, which can then be used to train smaller models. Attackers employed techniques like framing queries as translation requests to trick models into revealing internal thinking traces. This activity highlights escalating competition and concerns over intellectual property in the AI space.

Find the right AI tool for your business

Chat with Insta and get matched to the right tool in seconds.