Anthropic, a U.S.-based artificial intelligence company, has accused Chinese e-commerce conglomerate Alibaba Group of illicitly accessing Anthropic’s Claude language model. According to Anthropic, Alibaba created fake user accounts to circumvent access restrictions and conducted so-called “distillation attacks,” using Claude’s outputs to train Alibaba’s own models.
Anthropic says these actions involved using Claude’s responses as training material for Alibaba’s models. The company characterizes the behavior as unauthorized or in breach of access controls, highlighting concerns about access security and the use of large language model outputs.
Legal and geopolitical context
These allegations come as Alibaba this week sued the U.S. Department of Defense (the Pentagon), seeking removal from a blacklist of firms that are allegedly linked to the People’s Liberation Army (PLA).
The claim also arrives against a backdrop of broader U.S. policy aimed at reducing dependence on Chinese technology supply chains across sectors from raw materials to semiconductors. That effort, sometimes referred to as “Pax Silica,” seeks to bolster alternative sources. Several European governments and the European Union joined similar measures this week.
Why this matters
The dispute between Anthropic and Alibaba underscores the strategic and commercial risks tied to access to large language models and the reuse of their outputs. If Anthropic’s allegations are substantiated, the case could deepen legal disputes among tech companies, prompt stricter regulatory oversight, and strengthen measures to protect intellectual property and data usage norms.
Further developments, including evidence presented and legal proceedings, will determine how the matter unfolds and what precedents it may set.
