Mythos isn't crying wolf

Mozilla shares its receipts

While schools grapple with the Canvas ransomware attack during finals week, Anthropic’s Mythos is the real deal for IT security.

In April, Mozilla fixed 423 security bugs, more than the 258 they fixed in all of 2025.

In yesterday’s blog post, they revealed their new AI-based approach for rooting out security bugs. It found gaps that had survived for decades, including a 20 year old bug in their XSLT code.

Quick, use “Extensible Stylesheet Language Transformation” in a sentence. 😂

The clock is ticking. Security expert Alex Stamos estimates that we have 6 months before other frontier models reach Mythos-level capability, 12-18 before it is available in open-source models.

This is good news for Firefox users. Other software makers should follow Mozilla’s example and incorporate AI tools in their testing practices. And companies should make sure their software vendors are doing the same.