Anthropic's AI finds 22 security issues in Firefox
Anthropic's AI, Claude Opus 4.6, teamed up with Mozilla and uncovered 22 security issues in Firefox, including 14 classified as high-severity.
These made up nearly 20% of all critical bugs fixed in the browser earlier this year (February 2026).
Claude scanned Firefox's JavaScript engine
Claude started by scanning Firefox's JavaScript engine and spotted a dangerous bug within minutes.
It didn't stop there, producing 50 more unique crashing inputs and ultimately contributing to 112 submitted reports (including 22 confirmed vulnerabilities), including additional issues and crashing inputs that the team triaged and assessed for security impact.
Most flaws were patched in Firefox 148
Mozilla quickly patched most of these flaws in Firefox 148 (released in February 2026).
Still, Claude managed to pull off two basic hacks after spending $4,000 on API credits, showing just how powerful modern AI tools have become at finding real-world vulnerabilities.
AI's role in cybersecurity is expanding rapidly
AI like Claude is changing how software gets secured; it can now find, test, and even help fix bugs at scale.
Mozilla cooperated with Anthropic on triage and encouraged bulk submissions, facilitating the reporting and patching process.
For anyone into tech or cybersecurity, this is a glimpse at where things are headed next.