Anthropic's AI vulnerability hunter: finds what humans taught it, not more

Anthropic's security audit tool detects code vulnerabilities but only patterns it was trained on.

Published 12sem1 sourceNotable

≈ 18sSpeed

The fact

The system shows uneven performance on novel or underrepresented flaws.

This reveals limits to AI security tools' generalization capabilities.

Click the link to read an article on the topic:

Explore this topic

Auto-synthesis from 1 media source · identified on April 27, 2026