China has raised a red flag over a popular AI coding tool. On Wednesday, Chinese officials warned that Anthropic’s Claude Code has a security problem.
They say it could put user data at risk. This news comes at a time when the U.S. and China are locked in a growing rivalry over artificial intelligence.
China’s Warning
China’s Ministry of Industry and Information Technology made the announcement. Its cybersecurity threat platform found something troubling.
Officials say Claude Code has a “back-door” vulnerability. They call it a serious threat. So what does that mean in plain terms? The tool may send private data to an outside server.
And it can do this without asking the user first. According to the Chinese statement, it could include a user’s location and possibly identity.
Neither of those things should leave a device without permission.
Affected Versions
Chinese officials pointed to a specific range- versions 2.1.91 through 2.1.196. Anthropic’s own changelog helps fill in the timeline.
Those versions rolled out between April 2 and June 29; roughly a three-month window.
Anthropic has since moved on to a newer version is 2.1.204. Officials in China are urging users to act fast. They recommend uninstalling the affected versions. Or better yet, upgrade right away.
Anthropic’s Response
Anthropic responded once reporters reached out about the warning. Its explanation paints a very different picture than “security flaw.”
According to Anthropic, it was intentional. The company says it built this feature as an experiment earlier this year to protect its AI models from something called distillation.
Distillation is when one company tries to copy how another company’s AI thinks. They do this by studying its outputs closely, then they train their own model to mimic it.
Companies spend huge amounts of money building AI models. So they don’t want rivals getting a shortcut.
Anthropic also pointed to its official policy that blocks use by companies mostly owned by China-based organizations. In other words, the rule has been part of how Anthropic operates.
Technology
This isn’t the first clash between Anthropic and Chinese tech firms. Just last month, Anthropic accused Chinese firm Alibaba of trying to extract its AI capabilities.
Those capabilities aren’t even officially sold in China. Alibaba hasn’t publicly responded to that accusation. But the tension between the two companies is clearly building.
Interestingly, Claude Code has found its way into China anyway. Despite not being officially available there, locals have used workarounds.
Back in March, a developer from Xiaomi spoke at a state-organized forum. That person noted many people were already using Claude Code, official access or not.
Alibaba has now told employees to stop using Anthropic’s tools at work. That ban started July 10, according to reporting confirmed earlier this week.
