The already fragile state of US-China tech diplomacy just got a lot more complicated. China has set preconditions for upcoming AI negotiations with the United States while simultaneously lobbing accusations at Anthropic, the maker of the Claude AI model, in what is shaping up to be one of the most consequential standoffs in the brief history of artificial intelligence.
The dispute centers on allegations that Alibaba’s Qwen AI lab orchestrated a sweeping campaign to extract capabilities from Anthropic’s Claude model. Anthropic claims nearly 25,000 fraudulent accounts generated over 28.8 million unauthorized queries between April 22 and June 5, targeting Claude’s most advanced features in software engineering and autonomous reasoning.
The distillation allegations
On June 10, Anthropic sent a letter to US senators laying out what it described as the most extensive recorded exploitation of the Claude model. The technique in question, known as model distillation, involves using one AI system’s outputs to train a cheaper, smaller model.
The US government responded swiftly. The Commerce Department restricted access to Anthropic’s advanced AI models in mid-June, citing national security concerns.
Beijing fires back
China’s response has been pointed and multi-pronged. In July, Beijing accused the United States of pursuing “AI hegemonism” after sanctions were proposed in connection with the distillation accusations.
China’s Ministry of Industry and Information Technology also issued warnings about security vulnerabilities in Anthropic’s Claude tool.
Beijing has also threatened countermeasures to safeguard its interests. This tit-for-tat dynamic now threatens to derail bilateral AI safety talks scheduled for September. Those discussions were supposed to establish guardrails around AI development, covering everything from autonomous weapons to alignment research.
What’s actually at stake
China’s decision to set conditions for the talks signals that Beijing won’t simply show up and accept an American-defined agenda.
The Commerce Department’s restrictions on Anthropic’s own models add an ironic wrinkle. The US government is simultaneously trying to protect American AI from Chinese exploitation and restricting how American AI companies can deploy their most capable systems. Anthropic finds itself caught between two policy impulses: protect the technology, but also limit it.