Anthropic is removing a hidden tracking feature from Claude Code, its AI coding tool, after a wave of criticism from developers.
According to The Information's Juro Osawa (via Techmeme), the company is rolling back a covert feature designed to identify users based in China or affiliated with Chinese AI labs. Reporting from MSN says the code quietly tracked users' locations and their possible links to China, and that the outcry played out largely on social media, as The Decoder notes.
Anthropic's stated rationale, according to MSN, was anti-abuse enforcement, tied to national security concerns. Reporting there says the company has identified and suspended accounts it suspected of being tied to restricted users. In other words, Anthropic frames the mechanism as a way to keep certain foreign actors off its models rather than as general surveillance.
Critics saw it differently. Several outlets, including India Today and MSN, described the feature as "spyware" and accused Anthropic of monitoring users without telling them. The core complaint was not only what the code did, but that it was hidden — embedded in the tool and, as one MSN headline put it, concealed from the people using it.
Facing that backlash, Anthropic says it is reversing course and stripping the monitoring code out of Claude Code.
The episode matters because it shows how quickly the line between security enforcement and covert user tracking can blur — and how much trust an AI company can lose when developers discover their tools are watching them without disclosure.