China Rebukes Anthropic, Sets Terms for Key US-China AI Dialogue

China laid out conditions for artificial intelligence talks with the US, striking a strident tone ahead of a much-anticipated meeting while accusing American AI company Anthropic of bad behavior.

A Sunday post on Yuyuantantian, a social media account affiliated with Chinese state broadcaster CCTV, said the US must prove its AI companies are subject to the same safety, disclosure and audit rules before any “substantive” discussions with China can take place.

“A clear distinction must be drawn between genuine security threats and mere technological competition,” said the post, published by an account frequently used to signal official policy thinking. “This line must be drawn jointly by all participating parties,” it added.

Officials from the US and China are expected to hold talks on AI before Xi Jinping’s state visit to the US on Sept. 24 for a summit with Donald Trump. No date has been announced for that AI dialogue and Chinese officials last month were still seeking clarity on the agenda, in part to decide who to send, Bloomberg previously reported.

Tensions are mounting around the AI race between the world’s biggest economies. Underscoring that point, Yuyuantantian singled out Anthropic’s Claude for criticism, accusing that model of overstepping user data boundaries, engaging in covert monitoring and transmitting website domains without authorization.

“The problem is that America’s own frontier models have already developed in a distorted direction,” Yuyuantantian’s post said. “This means the negotiation is not simply a technical dialogue from the start, but a continuation of all the earlier problems: the US is trying to turn the ‘safety boundaries’ it has drawn into the default rules for the entire world.”

The US government should be guarding security boundaries around AI, yet it has “repeatedly retreated” from doing so, according to the post titled “Anthropic Has Contracted the American Disease.”

Anthropic and the US State Department did not immediately respond to requests for comment sent outside US office hours.

The Yuyuantantian post said the US approach to AI reflects a broader effort to contain China’s technological rise. “The ‘control’ proposed by the US is, in essence, an attempt to make China accept an order partly defined by American companies,” it added.

Chinese officials are growing anxious that Anthropic’s Mythos could be wielded against their country, Bloomberg previously reported. They see the potential for models like Mythos to be used as an offensive weapon, and question why Anthropic is denying China access to the systems for normal purposes, people familiar said.

Anthropic blocked its services from Chinese-controlled companies last year, citing US national security concerns. But that does not seem to have effectively stopped Chinese companies from using its products.

The firm, along with OpenAI and Alphabet Inc.’s Google, has expressed concern about its models being used in China. Anthropic has also accused Alibaba Group Holding Ltd. of “illicitly” accessing Claude using thousands of fraudulent accounts, an allegation the Chinese tech giant has not responded to.