Anthropic’s Claude finds 22 Firefox security bugs, Mozilla fixes them

In short: Anthropic says its AI model Claude Opus 4.6 found 22 security vulnerabilities in Firefox, and Mozilla fixed them in Firefox 148.

What happened

Anthropic worked with Mozilla for two weeks to look for security problems in Firefox. Anthropic said its Claude Opus 4.6 model found 22 separate vulnerabilities, and 14 were labeled high-severity.

The effort started after Anthropic’s Frontier Red Team contacted Mozilla. They said Claude had already found more than a dozen real, repeatable bugs in Firefox’s JavaScript engine, which is the part of the browser that runs code on many websites.

Anthropic said one of the first findings happened fast, within about 20 minutes. It was a “use after free” bug, which is a memory mistake where the program keeps using a piece of memory after it has been thrown away (like writing on a sticky note that is already in the trash). In some cases, bugs like this can let an attacker overwrite data with harmful instructions.

Mozilla said it fixed all 22 security issues in Firefox version 148, released February 24, 2026. Anthropic also reported about 90 additional non-security bugs, such as crashes and logic errors.

Why it matters

Most people only notice browser security when something goes wrong. Fixing bugs like these helps reduce the chance that someone can break into your browser through a malicious website. Mozilla also noted that modern browsers have several layers of protection, so an attacker often has to combine multiple flaws to cause real harm.

Source: TechCrunch AI