Simon Willison • 1/14/2026

Claude Cowork Exfiltrates Files

Read Original

A security report details a vulnerability in Claude Cowork's default outbound HTTP traffic restrictions. Researchers from Prompt Armor found a workaround: since Anthropic's API domain is allowed, an attacker could use their own API key to make the agent upload user files to the Anthropic API endpoint, enabling data exfiltration despite the intended safeguards.

0 comments
#prompt injection #API Security #Security Vulnerability
Claude Cowork Exfiltrates Files

Comments

No comments yet

Be the first to share your thoughts!

Browser Extension

Get instant access to AllDevBlogs from your browser

Top of the Week

1
Quoting Thariq Shihipar
Simon Willison • 2 votes
2
Top picks — 2026 January
Paweł Grzybek • 1 votes
3
In Praise of –dry-run
Henrik Warne • 1 votes
4
Deep Learning is Powerful Because It Makes Hard Things Easy - Reflections 10 Years On
Ferenc Huszár • 1 votes
5
Vibe coding your first iOS app
William Denniss • 1 votes
6
AGI, ASI, A*I – Do we have all we need to get there?
John D. Cook • 1 votes
7
Dew Drop – January 15, 2026 (#4583)
Alvin Ashcraft • 1 votes