In a review of our cybersecurity evaluation transcripts, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations. Below we describe what happened, how it happened, and what we’re changing. We encourage other AI labs to perform similar reviews.
The Verdict
ClassificationUnsure
ConfidenceHigh confidence
Analyzedtext
Community Verdict
Sign in to vote
Be the first to vote on this assessment.
Embed Badge
Add this badge to your site to show the AI classification for this content.
[](https://real.press/content/e19cf4c6-5d64-4124-8c25-d207f65daffd)