Posted in Vulnerability Management
About this dashboard
Gheorghe B. — Senior Software Engineer
Quite like this detail in Anthropic’s vulnerability dashboard.
Some of the numbers went down in the latest update.
They found duplicates and withdrew some findings, then explained the correction. The ledger keeps the history and separates the model’s severity rating from the security firm’s and the maintainer’s.
Honestly, this is the part of AI security research I want more of.
Finding thousands of things is interesting. Showing which ones survived scrutiny, and leaving the corrections visible, is considerably more useful.
https://red.anthropic.com/2026/cvd/about/

4 reactions · 0 comments