AI · 237 days ago

AI finds 22 Firefox flaws; generated exploits in two cases

Anthropic’s Claude Opus 4.6 found 22 vulnerabilities in Mozilla Firefox during a two-week January 2026 test. Mozilla rated 14 as high severity and most issues were fixed in Firefox 148. After hundreds of attempts and about $4,000 in API spend, Claude produced working exploit code for two bugs.

Timeline

• Mar 7 Reported by The Hacker News
• Mar 6 Reported by TechCrunch Security

Sources

2 sources covering this story

• The Hacker News Mar 7

  Anthropic Finds 22 Firefox Vulnerabilities Using Claude Opus 4.6 AI Model

  Anthropic’s Claude Opus 4.6 AI found 22 Firefox vulnerabilities, including 14 high severity, helping Mozilla patch flaws in Firefox 148.

• TechCrunch Security Mar 6

  Anthropic's Claude found 22 vulnerabilities in Firefox over two weeks | TechCrunch

  In a recent security partnership with Mozilla, Anthropic found 22 separate vulnerabilities in Firefox — 14 of them classified as "high-severity."

Part of the PlainSec briefing for 2026-03-08

Editions

• Briefing for 2026-03-08

Related stories

• OpenAI agent finds 792 critical and 10,561 high-severity flaws