tech
May 7, 2026
How Anthropic’s Mythos has rewritten Firefox’s approach to cybersecurity
Security researchers at Mozilla say Anthropic's Mythos has unearthed a wealth of high-severity bugs in Firefox.
TL;DR
- Anthropic's Mythos AI model has discovered thousands of high-severity software vulnerabilities.
- Mozilla's Firefox security team has seen a dramatic increase in bug discoveries attributed to Mythos.
- The AI has found bugs that have existed in the code for over a decade, including complex sandbox vulnerabilities.
- Agentic AI systems can now assess their own work, overcoming previous issues with false positives and low-quality reports.
- In April 2026, Firefox shipped 423 bug fixes, a substantial increase from 31 a year prior.
- Human engineers are still necessary for writing and reviewing code patches, as AI-generated code is not yet deployable.
- The impact of AI on the cybersecurity balance of power is still unfolding, with potential benefits for both attackers and defenders.