Google AI Helps Chrome Fix 1,072 Vulnerabilities in a Month, Surpassing Two-Year Total

Google's Chrome security team recently disclosed that using large language models (LLMs), it fixed 1,072 security vulnerabilities in Chrome versions 149 and 150 released in June, surpassing the total of 1,036 vulnerabilities fixed in the previous 23 versions over nearly two years. Among them, AI successfully discovered and fixed a 13-year-old sandbox escape vulnerability that allowed local file access. This achievement has been recognized by security experts, marking AI's huge potential in software engineering security and sparking discussions about developers' survival space.

Confirmed

• Scale of fixes: AI-assisted fixes in June addressed 1,072 security vulnerabilities, exceeding the total of 1,036 fixed in the previous 23 versions over nearly two years.
• Scope of application: AI (LLM) is deeply integrated into the entire vulnerability lifecycle, from discovery (e.g., improving fuzzing coverage, collaborating with Project Zero on Naptime, and with DeepMind on Big Sleep) to patching.
• Notable case: AI discovered and fixed a 13-year-old sandbox escape vulnerability that allowed local file access.
• Engineering architecture: According to Pawel Huryn, Google built a knowledge base integrating all known Chrome CVE vulnerability databases and complete Git commit history to support AI fixing work.

Unconfirmed

• Some posts (e.g., m3) mention 1,442 vulnerabilities fixed, differing from the official 1,072; this may involve different versions or statistical scopes and needs further verification.

Why it matters

• Introducing AI into the full lifecycle of software vulnerability discovery and fixing breaks through the limits of human security experts' energy, greatly improving the efficiency of code auditing and security maintenance, and providing a new paradigm for security defense across the software industry.

2026-07-31 ~ 2026-08-01 · 14 related posts

• Google Chrome Uses AI to Accelerate Bug Discovery and Fixing, Over 1000 Fixes Applied — laparisa · 2026-07-31
• [source] Google Fixed More Chrome Bugs in June Using AI Than in the Previous Two Years Combined — ControlCAD · 2026-07-31
• AI Accelerates Bug Discovery: Chrome Patches Record 1,442 Flaws — gnukeith · 2026-07-31
• [source] AI Uncovers 13-Year-Old Chrome Security Bug — petrusenko_max · 2026-08-01

10 near-duplicate retellings: laparisa · dyn___ · laparisa · luisdans · PawelHuryn · PawelHuryn · laparisa · steren · VraserX · zainhas