China Warns Anthropic and Sets Preconditions for U.S.-China AI Dialogue
nashnova research
A Chinese state-media-linked account singled out Anthropic's Claude for data overreach, demanding the US first prove its own AI firms face equivalent oversight — raising the stakes for AI talks ahead of the planned Xi-Trump summit.
Who spoke up, and what did they say?
The message came from Yuyuan Tantian, a social-media account run by China Media Group (the state broadcaster) and widely seen as a channel for official policy signals.
The article, titled *Anthropic Has Caught the American Disease*, accused Claude of collecting user data beyond its scope, conducting covert surveillance, and transmitting website domains without authorization.
This means → This is not ordinary media commentary — it is a pre-negotiation signal sent through a semi-official outlet.
What conditions is China setting?
The core demand: the US must first prove its own AI companies are subject to the same safety, disclosure, and audit rules it expects of China before bilateral AI talks can make "substantive" progress.
The article framed US-led AI "safety boundaries" as a tool for imposing American corporate standards on the world, and accused Washington of "repeatedly retreating" from enforcing those very boundaries.
In plain terms = Beijing's message is: if you want to set AI rules with us, regulate your own companies first — you cannot be both referee and player.
Why has Anthropic become the target?
Anthropic blocked Chinese-controlled entities from accessing Claude last year, but the block reportedly did not fully hold — Chinese firms continued using its products.
Anthropic also accused Alibaba of "illegally" accessing Claude through thousands of fake accounts; Alibaba did not respond.
Chinese officials are wary of Anthropic's Mythos model, fearing it could serve as an offensive tool against China, and question why Anthropic cites US national security to deny China normal access.
This reflects a dual trigger: Anthropic restricted Chinese access on a commercial level and became, in Beijing's eyes, the embodiment of America's "security narrative."
Can AI talks still happen before the summit?
According to Bloomberg, US and Chinese officials had planned a dedicated AI dialogue before Xi Jinping's September 24 visit to meet Trump, but no date has been announced.
Chinese officials were still seeking clarity on the agenda last month to decide what level of delegation to send.
This means → By publicly linking Anthropic to preconditions for the AI dialogue, Beijing has added another hurdle to the table — even if the summit goes ahead, whether the two sides can bridge their fundamental divide on AI governance remains the biggest open question.
市场有风险，内容仅供研究参考，不构成投资建议。