Firefox finds a slew of new bugs with Claude's help • The Register

Firefox has made significant strides in security thanks to Anthropic's AI bug-hunting capabilities, but Mozilla engineers have uncovered a more mundane culprit behind many browser crashes: faulty hardware. The discovery highlights both the promise of AI-assisted security and the persistent challenges of hardware reliability in modern computing.

AI-Assisted Security Breakthrough

Several weeks ago, Anthropic approached Mozilla with a new AI-based vulnerability detection system that promised to revolutionize how the Firefox team identifies security flaws. Unlike previous AI-assisted bug detection attempts that yielded mixed results, this system proved remarkably effective.

Mozilla engineers Brian Grinstead and Christian Holler described the collaboration as transformative. Within hours of implementing the AI system, platform engineers began landing fixes across the browser codebase. The results were impressive: 14 high-severity bugs discovered and 22 CVEs (Common Vulnerabilities and Exposures) issued, all now patched in the latest Firefox version.

Anthropic achieved this feat using its Claude Opus 4.6 model, which even generated a working exploit for one of the vulnerabilities (CVE-2026-2796). The company emphasized that this exploit only worked in a testing environment where security features had been intentionally removed. As security researchers Evyatar Ben Asher, Keane Lucas, Nicholas Carlini, Newton Cheng, and Daniel Freeman explained, "Claude isn't yet writing 'full-chain' exploits that combine multiple vulnerabilities to escape the browser sandbox, which are what would cause real harm."

However, Anthropic warned that this limitation may not last long. "Looking at the rate of progress, it is unlikely that the gap between frontier models' vulnerability discovery and exploitation abilities will last very long," the company stated. This raises important questions about future safeguards as AI capabilities continue to advance.

The Hidden Hardware Crisis

While AI is helping secure Firefox's software, Mozilla engineer Gabriele Svelto has identified a more fundamental problem affecting browser stability: hardware failures. In a recent Mastodon post, Svelto revealed that approximately 10 percent of Firefox browser crashes can be attributed to bit flips – unintentional changes in memory caused by cosmic rays, Rowhammer attacks, or more commonly, flawed electronic components.

Svelto's research uncovered startling statistics. In just one week, Mozilla received about 470,000 crash reports from Firefox users who opted into crash reporting. Of these, approximately 25,000 appeared to be potential bit flips – "one crash every twenty potentially caused by bad/flaky memory," Svelto noted. He emphasized that this is a conservative estimate, suggesting the real number could be at least twice as high.

When crashes caused by resource exhaustion (like running out of memory) are excluded, the proportion of hardware-related crashes rises to about 15 percent. Svelto stressed that while his research focuses on computers and phones, these issues affect every device, from routers to printers.

This hardware reliability problem isn't new. Google researchers discovered in 2009 that DRAM error rates in data centers were "orders of magnitude higher than previously reported, with 25,000 to 70,000 errors per billion device hours per Mbit and more than 8 percent of DIMMs affected by errors per year."

The Broader Implications

The dual discoveries – AI's growing security capabilities and hardware's persistent unreliability – paint a complex picture of modern computing. On one hand, AI is proving to be an invaluable tool for identifying and patching software vulnerabilities faster than ever before. On the other hand, no amount of software sophistication can overcome fundamental hardware flaws.

Bit flips represent a particularly challenging problem because they're beyond the control of software developers like Mozilla. While techniques exist to mitigate some hardware errors, the scale of the problem suggests that device manufacturers need to prioritize component quality more seriously.

Svelto's findings also raise questions about how we measure software reliability. If a significant portion of what we consider "crashes" are actually hardware failures, our understanding of software stability may need revision. This could impact everything from user experience metrics to how we develop and test software.

As AI continues to advance in its ability to find and exploit vulnerabilities, the industry faces a paradox: we're getting better at securing software just as we're discovering that hardware reliability remains a significant, unsolved challenge. The path forward likely requires both continued AI innovation and renewed focus on hardware quality across the entire technology ecosystem.

For Firefox users, the good news is that the AI-assisted security improvements are already in place, making the browser more secure than ever. The bad news is that if you're experiencing crashes, the problem might not be Firefox at all – it could be your device's memory or other components failing in ways that no software update can fix.