Anthropic removes hidden telemetry feature from Claude Code
Anthropic has removed an undocumented telemetry feature embedded in its Claude Code developer tool after users and security researchers flagged the behavior. The transmissions were not surfaced in release notes or public-facing privacy materials, and Anthropic confirmed the change in response to community reports.
What was happening
Users noticed that Claude Code appeared to be sending activity data back to Anthropic without clear disclosure in the tool’s documentation. Developers using Claude Code, a command-line interface for working with Anthropic’s Claude models, began comparing notes online. The shared observations suggested background logging of user interactions, though the exact scope and contents of the data were not independently confirmed in the source reporting available.
How did Anthropic respond?
After the issue gained visibility, Anthropic acknowledged the problem and removed the feature. The company did not, in the available reporting, provide a detailed explanation of how long the undocumented telemetry had been active, what specific data it collected, or how many users were affected.
Why does it matter for AI tool transparency?
The incident highlights a recurring tension in AI development: tools that process sensitive code, business logic, and personal data often run locally, giving users a sense of privacy. When background data flows exist without clear disclosure, that trust is put at risk.
- Open documentation of telemetry is now treated as a baseline expectation by many developer teams.
- Security audits of AI tooling increasingly focus on outbound network requests, not just local behavior.
- Vendors face pressure to align privacy disclosures with what their software actually does in practice.
What should developers do?
Teams using Claude Code, or any AI-powered CLI, are advised to monitor network traffic from their development environments and keep tooling up to date. Reviewing release notes and changelogs before updating, and flagging undocumented behavior to vendors, remains a practical defense.
FAQ
What did Anthropic remove from Claude Code?
Anthropic removed an undocumented telemetry feature embedded in Claude Code that was sending activity data back to the company without clear disclosure in the tool’s documentation or release notes.
Was the Claude Code telemetry disclosed before it was removed?
No. According to the available reporting, the feature was not described in the public-facing privacy materials or release notes at the time users flagged it.
What should developers do in response?
Developers are advised to monitor network traffic from their development environments, keep tooling up to date, review release notes and changelogs before updating, and flag undocumented behavior to vendors.