Anthropic quietly patches vulnerabilities in Claude Code
Security researcher Aonan Guan found a second critical flaw in five months in Claude Code’s network sandbox. The bug allowed bypassing filters and sending sensitive data like GitHub tokens and cloud credentials to arbitrary external servers.
Anthropic fixed the issue in version 2.1.88 but did not publish a CVE or official security notice. The company considers the in-repo fix in sandbox-runtime sufficient, but experts criticize this approach.
The main risk is a false sense of security: users think their data is protected by the sandbox, but network boundaries are effectively absent. Experts advise treating AI agents like employees with limited access rights, not regular software.
Up next
Claude Code update v2.1.146
The /simplify command in Claude Code is now /code-review, with new depth settings like ‘high’ for thorough analysis. This is more than a rename; it’s a repositioning of the tool as a stricter, full-fledged team collaborator rather than just a quick assistant.
Related stories
- Apiiro finds Claude Code secrets in 1,122 GitHub files
- Email leak stopgap in Claude Code checks only shell calls
- Claude Code agent sat on a Pentagon network for two months
- Backdoored Claude Code builds were limited to Cursor
- Claude Code sends file hashes to Anthropic even on DeepSeek
- RuntimeWire: Claude Code puts staff bypass in Accept edits
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.