Anthropic Accuses Alibaba of Illicitly Extracting Claude AI

Anthropic has accused Chinese tech giant Alibaba of illicitly obtaining access to its Claude AI models and extracting their capabilities, according to reports published simultaneously by Reuters, Bloomberg, and the Financial Times. The allegations represent one of the more serious intellectual property disputes to emerge from the fast-moving AI industry, and they arrive at a moment of heightened tension between American AI companies and Chinese technology firms.

What Anthropic Is Alleging

According to the reports, Anthropic claims that Alibaba gained access to Claude through means that violated the company's terms of service. The specific method of extraction has not been fully detailed in public filings, but sources familiar with the matter indicate the concern centers on systematic querying or scraping designed to reproduce model behavior or underlying capabilities. Anthropic appears to be treating this not as a minor breach but as a deliberate attempt to replicate proprietary AI technology.

The timing adds context to recent regulatory moves in the United States. The Biden and Trump administrations have both taken steps to limit foreign access to advanced American AI systems. Earlier this year, Anthropic moved to disable its top AI models under a US foreign access order, a decision that reflected growing government pressure on frontier AI labs to restrict who can use their most capable systems. Anthropic's accusations against Alibaba fit into that broader pattern of concern about technology transfer.

Alibaba's alleged actions represent a direct attempt to circumvent the access controls and legal agreements that govern the use of Claude.Anthropic, via Reuters

Alibaba's Position and the Competitive Backdrop

Alibaba has not publicly accepted the accusations. The company has been investing aggressively in its own AI capabilities, most visibly through its Qwen model series. Alibaba's Qwen3.7-Max model recently drew attention for running autonomously for 35 hours and supporting Claude Code workflows, a signal of how seriously the company is competing in the agentic AI space. Critics of Anthropic's position may argue that competitive pressure can lead to overzealous IP claims, though the specificity of the allegations suggests Anthropic has collected evidence before going public.

For Anthropic, the stakes are significant. The company has spent years and hundreds of millions of dollars developing Claude's model family, and any systematic extraction of those capabilities by a competitor would undermine that investment directly. It would also raise difficult questions about how AI companies can protect their models from what amounts to a form of competitive intelligence gathering at scale. Traditional software IP law was not designed with large language models in mind, and cases like this one may help establish new legal precedents.

What Comes Next

It is not yet clear whether Anthropic intends to pursue legal action or whether the accusations are part of a broader regulatory or diplomatic effort to curb Chinese firms' access to American AI systems. The involvement of major outlets like Reuters and the Financial Times at the same moment suggests a coordinated disclosure, which could indicate that formal legal proceedings or regulatory filings are imminent.

The dispute is unlikely to remain an isolated incident. As frontier AI models become more commercially valuable, attempts to extract or replicate their capabilities will almost certainly increase. The industry will be watching closely to see how Anthropic pursues this case and whether it sets any meaningful standard for how model developers can defend their intellectual property in a global market where enforcement is far from straightforward.