China Warns Claude Code Versions 2.1.91-2.1.196 Pose Back-Door Risk
Updated
Updated · CNBC · Jul 8
China Warns Claude Code Versions 2.1.91-2.1.196 Pose Back-Door Risk
3 articles · Updated · CNBC · Jul 8
Summary
China's industry ministry said Anthropic's Claude Code can transmit sensitive data to a remote server without user consent, exposing information such as location and identity.
The warning covers versions 2.1.91 through 2.1.196, released from April 2 to June 29, and users were told to uninstall or upgrade; Anthropic's latest listed version is 2.1.204.
The alert lands as Chinese companies and developers have been using the U.S. coding tool despite its unavailability in China, with a Xiaomi AI developer saying in March that many locals used Claude Code.
Alibaba already ordered employees to stop using Anthropic tools for work from July 10, after Anthropic last month accused the Chinese company of trying to extract its AI capabilities.
As China flags a 'backdoor' in a top U.S. AI, is this a real cyber threat or a tech war tactic?
If AI assistants can secretly leak data, can any company truly guarantee its software is secure in the AI era?
With both nations using AI to find software flaws, who is gaining the ultimate advantage in the cyber arms race?
2026 Claude Code Scandal: Anthropic’s Covert China Ban, Alibaba’s Response, and the Future of AI Security and Transparency
Overview
In early July 2026, security researchers uncovered and shared evidence on Reddit and GitHub revealing that Anthropic’s terms of service banned Chinese companies and other so-called adversarial nations from using their advanced AI models. This public disclosure quickly led Alibaba to ban Anthropic’s products from its ecosystem to protect its data and comply with national regulations. The incident highlighted the risks of relying on foreign AI providers with restrictive policies and triggered a strong response from Chinese authorities, intensifying the push for technological self-reliance and stricter oversight of foreign technology in China.