China's industry ministry said on Wednesday that it had identified a serious security "backdoor" risk in Anthropic's AI coding tool, Claude Code. The warning lands where power always does: at the point where workers, developers, and companies are told what tools they can use, what data can move, and who gets to decide.
The National Vulnerability Database, in a statement posted on its WeChat account, said Claude Code contains a built-in monitoring mechanism capable of transmitting sensitive information, including users' geographic location and identity-related identifiers, to remote servers without users' consent. That's the kind of arrangement that turns a coding tool into a surveillance channel, with ordinary users left to absorb the risk while institutions issue the orders.
Who Gets Watched
The warning applies to Claude Code versions 2.1.91 through 2.1.196. The database advised organizations and users to immediately review affected systems and either uninstall the impacted versions or upgrade to the latest secure release in which the alleged backdoor code has been removed. It also urged organizations to tighten controls on external network access for development tools and strengthen traffic monitoring on core business networks to prevent the unauthorized transfer of sensitive data.
That language is all control, all the time. Review systems. Uninstall tools. Tighten access. Monitor traffic. The people actually using the software don't get a say in the architecture of the system; they get instructions after the fact, once the apparatus has already decided the terms.
China's Alibaba has already banned employees from using Claude Code at work after the tool drew scrutiny for features that can help identify China-linked users, Reuters reported last week. Workers don't get freedom here. They get a corporate prohibition, handed down from above, justified in the name of security and enforced through the workplace hierarchy.
Anthropic said what was being described as a "backdoor" was an experimental anti-abuse mechanism, and that access to Claude was not permitted in China. The company’s defense doesn't erase the basic fact that access, permission, and surveillance all sit in the hands of institutions with the power to set the rules.
What the Authorities Are Moving
Separately, Breakingviews reported that Chinese authorities held meetings with top tech firms over the past month about potentially restricting overseas access to China's most advanced AI models, including models not yet released. Participants reportedly included Alibaba, ByteDance and Knowledge Atlas. The report said such curbs could trigger cascading costs for businesses.
That is the other side of the same machine. On one end, regulators warn about data leaving the country through foreign tools. On the other, they consider limiting who can reach domestic models beyond China's borders. Either way, the decision-making stays concentrated at the top, while businesses and users are left to live with the consequences.
The meetings over the past month show how quickly these controls move from abstract policy to direct pressure on major firms. Alibaba, ByteDance and Knowledge Atlas reportedly sat in the room while authorities discussed restrictions that could reshape access to advanced AI systems before some of them are even released.
Who Pays for the Controls
The report said such curbs could trigger cascading costs for businesses. That's the bill. Not for the people making the rules, but for the firms and workers downstream who have to absorb the disruption, the compliance burden, and the narrowed access.
The two developments point to tighter scrutiny around AI tools and model access in China, with regulators focusing on security controls, data transfer risks and the possible overseas reach of domestic AI systems. In plain terms, the state and the companies around it are drawing harder lines around who can use what, where information can go, and how much autonomy anyone gets before the next warning, ban, or meeting arrives.