Anthropic is facing criticism after reporting surfaced about a hidden Claude tracking experiment that monitored some Chinese users. The disclosure jars with Anthropic’s public positioning against surveillance uses of AI.
An engineer described the tracker as an experiment that is now over, according to Ars Technica. The episode raises questions about how AI labs test abuse-prevention systems and what notice users should receive.
For a company that emphasizes safety and trust, even narrowly scoped monitoring can become a reputational issue if users believe the boundary was unclear.