Google Uses AI to Fix More Chrome Bugs in June Than Past Two Years
Original: Google fixed more Chrome bugs in June than over the past two years, thanks to AI
Why This Matters
AI-driven vulnerability discovery at scale signals a fundamental shift in how major browsers can maintain security at speed.
Google's Chrome Security Team announced on July 30, 2026, that AI-powered tools using LLMs—including Gemini and DeepMind's Big Sleep agent—helped discover and fix more Chrome vulnerabilities in June 2026 alone than in the previous two combined years, dramatically accelerating the security patch lifecycle.
Google's Chrome Security Team revealed that large language models (LLMs) are now central to its vulnerability discovery and remediation pipeline. The team has progressively scaled AI use since 2023, when LLMs were first applied to fuzzing. In 2024, Project Zero's 'Naptime' gave LLMs specialized research tools, and in 2025, the 'Big Sleep' agent—developed with DeepMind—successfully identified bugs in Chrome's V8 JavaScript engine and graphics stack.
In early 2026, the team built an agent harness using Gemini to scan the broader Chrome codebase with higher efficiency and fewer false positives. One notable find: a sandbox escape vulnerability that had existed undetected in the codebase for over 13 years. Subsequent improvements included multi-model interoperability (both open-weights and proprietary), a Chrome-specific knowledge base incorporating all historical CVEs and Git history, SECURITY.md files to help models understand trust boundaries, a dedicated 'critic' agent for threat modeling, and repeated model runs to account for non-determinism. The cumulative result: more bugs fixed in June 2026 than over the prior two years combined.