AI model uncovers hundreds of Firefox flaws during testing.

Mozilla said Anthropic’s Claude Mythos identified 271 vulnerabilities in Firefox, with more than 40 addressed in the browser’s latest release and three credited directly to the model. Many of the remaining issues appear to be lower‑severity bugs or hardening fixes that do not qualify for public CVEs. Mozilla emphasized that all of the flaws could also have been found by top human researchers, countering claims that AI will reveal entirely new classes of vulnerabilities. Early testing by Palo Alto Networks showed Mythos can chain lower‑severity bugs into critical exploits, prompting warnings that advanced cybersecurity‑focused AI models will soon introduce new enterprise risks.

Read more:

https://www.securityweek.com/claude-mythos-finds-271-firefox-vulnerabilities/