Anthropic Alleges Distillation Attacks by Chinese Labs

Anthropic has released a report alleging persistent distillation campaigns by China-based AI companies, naming Alibaba, Moonshot AI, and DeepSeek. Distillation is a technique in which a less capable model is trained using the outputs of a more advanced one, effectively transferring capabilities without the original developer's involvement. Anthropic frames the activity as a violation of its terms of service and a growing threat as competition in the AI industry intensifies. According to the report, the campaigns have escalated in recent months, suggesting a coordinated and sustained effort rather than isolated incidents. The allegations touch on some of the most contested issues in AI: model access, intellectual property, and national competitiveness. Distillation sits in a gray area. It is a standard technique used legitimately within companies to build smaller, cheaper models. But when applied to a competitor's outputs without permission, it raises questions about consent, fair competition, and the value of expensive training runs. For Anthropic, the report is both a technical disclosure and a strategic move, signaling that it is monitoring misuse of its systems and willing to name alleged violators publicly. The named companies have not yet responded in detail. The dispute highlights how unresolved the rules of the AI era remain, with no clear international framework governing how model outputs can be used, and how quickly tensions can escalate when commercial and geopolitical interests collide.