Anthropic Caught Secretly Tracking Chinese Claude Users, Contradicting Its Anti-Surveillance Stance

Anthropic Caught Secretly Tracking Chinese Claude Users, Exposed by Security Researcher

AI company Anthropic has removed a covert tracker that was secretly monitoring Chinese users of Claude Code, after a security researcher publicly exposed the hidden code and condemned the spyware-like behavior as a "serious violation of user trust."

The story, first reported by Ars Technica and picked up by Slashdot, has drawn widespread criticism given Anthropic's longstanding reputation as a privacy-conscious, safety-focused AI lab.

Researcher Stumbles Upon Hidden Tracking Code

Last week, a web developer going by the handle Thereallo was investigating Claude Code's privacy practices when they made a startling discovery: Anthropic had embedded user-tracking code targeting Chinese users using a technique known as prompt steganography.

Prompt steganography involves disguising instructions or data as seemingly innocuous content — effectively hiding information in plain sight so that ordinary users are unlikely to notice it, while the code continues to execute covert functions in the background.

A Direct Contradiction of Anthropic's Public Stance

What makes the incident particularly striking is that Anthropic has consistently positioned itself as a champion of user privacy and opponent of surveillance as part of its public brand identity. The discovery that the company was covertly monitoring users in a specific geographic region has not only raised serious questions about its integrity, but also reignited broader industry debate about transparency and privacy protection in AI products.

Anthropic Acts Quickly — But Stays Silent on Details

Facing mounting public pressure, Anthropic moved swiftly to remove the tracking code following the researcher's disclosure. However, the company has not yet issued a detailed public statement explaining the nature or purpose of the tracking, what data was collected, how long the tracker had been active, or why it was implemented covertly rather than disclosed in Anthropic's privacy policy.

A Warning Sign for AI Industry Transparency

The incident has renewed scrutiny of data collection practices across the AI industry. As AI-powered developer tools become increasingly ubiquitous in global developer communities, users frequently have little visibility into what information these tools collect in the background and how it is used.

Analysts note that this case serves as a stark reminder that AI companies — even those that market themselves on safety and ethics — still have significant room for improvement when it comes to meaningful transparency in their privacy practices.

Original reporting: Ars Technica / Slashdot

原文來源： 查看原文