Mozilla: AI-powered bug detection produces very few false positives

Mozilla says its AI-driven security workflow found 271 real Firefox vulnerabilities, including long-standing bugs and sandbox escapes.

The system reportedly produced almost no false positives, making it more reliable for large-scale bug discovery than earlier AI approaches.

Mozilla credits stronger models and a test harness that lets models run targeted checks and verify results before human review.

The success suggests AI could become a practical tool for faster security fixes and broader code testing across software teams.