Anthropic Deployed Spy Software on China-Based Claude Code Users in March
Updated
Updated · The Washington Post · Jul 6
Anthropic Deployed Spy Software on China-Based Claude Code Users in March
2 articles · Updated · The Washington Post · Jul 6
Summary
March marked a covert step by Anthropic: it quietly installed software to monitor China-based users of Claude Code, its coding chatbot.
Anthropic’s apparent aim was to identify Chinese rivals it suspected of “distilling” Claude’s knowledge — forcing the chatbot to act as a tutor for competing AI models.
The report frames the move as part of a broader U.S.-China struggle over AI secrets, with American tech firms alleging Chinese competitors are extracting know-how from their chatbots to accelerate development.
With China's AI user base now surpassing America's, has the U.S. strategy to contain its tech rival already failed?
When AI firms secretly spy on users to stop data theft, where is the line between security and corporate espionage?
Is 'distilling' a rival's AI a legitimate shortcut for innovation or is it simply high-tech industrial espionage?
Anthropic’s Hidden Tracking in Claude Code: 25,000 Chinese Accounts, Global Bans, and the Future of AI Transparency
Overview
In March 2026, a covert tracking feature called 'screening logic' was discovered in Anthropic's Claude Code, which had been secretly integrated without user disclosure. This hidden mechanism was a response to intense pressures on AI vendors, such as strict export-control regulations and competition from Chinese labs. However, instead of solving compliance issues, the secret feature led to public backlash and major consequences, including Alibaba banning Claude Code in its offices. The incident highlighted how undisclosed features can erode user trust and cause significant commercial damage, emphasizing the need for transparency and ethical practices in AI development.