Anthropic’s Claude AI model identified software bugs that conventional testing had not caught, according to a report published by Interesting Engineering. The work was described as a new approach to finding bugs in critical code.
The finding applies to AI-assisted formal verification in software security. The report did not describe independent verification of the results.
Read Full Article: https://www.naturalnews.com/2026-10-25-claude-finds-bugs-survived-software-test-runs.html
