tech
May 9, 2026
A company tested Claude Mythos Preview. It says the AI found a bug that existed for 20 years
Mozilla says Anthropic's unreleased Claude Mythos AI found security flaws that automated testing systems missed for years.
TL;DR
- Mozilla used Anthropic's unreleased Claude Mythos Preview AI to find security flaws.
- The AI model discovered hundreds of issues, including one that had existed for 20 years.
- This AI-assisted effort contributed to fixes for 423 security bugs in Firefox's April releases.
- Previous AI models generated incorrect solutions, but newer models can navigate complex codebases.
- Claude Mythos is not publicly released due to potential national security and online crime risks.
- Mozilla plans to integrate AI analysis further into Firefox's development pipeline.