Google Patches 1,072 Chrome Flaws in One Month, Fueled by AI Bug-Hunting Surge

Google patched a staggering 1,072 security vulnerabilities in its Chrome browser across two milestone releases shipped in June, a figure that eclipses the total number of flaws fixed across the previous 23 versions combined. The explosion in bug discoveries, detailed in a white paper published Thursday, is largely attributed to an internal AI-driven security overhaul that the company says has transformed vulnerability detection into an industrial-scale operation.

The fixes were distributed across Chrome 149 and Chrome 150, both released last month. By comparison, the 23 stable channel updates Google shipped over the prior two years contained a cumulative 1,036 security patches. The latest version, Chrome 151, released Wednesday, resolved an additional 370 flaws, 349 of which were flagged by Google's own internal tools. Seven of those vulnerabilities were rated critical.

"Large language models have fundamentally shifted the economics of cybersecurity, transforming vulnerability discovery into an automated, industrial-scale operation," Doug Turner, Chrome's director of engineering, said in a statement. "By applying models like Gemini, we are preemptively fixing vulnerabilities, outpacing our adversaries and making Chrome safer with every update."

The surge is not a sign of deteriorating code quality, Google emphasized, but rather a reflection of how effectively AI systems are now scanning the browser's massive codebase. One of the most significant discoveries made by these tools was a critical sandbox escape flaw in the Navigation component, tracked as CVE-2026-3545 with a CVSS score of 9.6. The bug, which could allow a compromised renderer process to trick the browser into reading local files from a user's system, had remained undetected in Chrome's source code for more than 13 years. It was patched in March.

That vulnerability was unearthed by an agent harness leveraging Google's Gemini models, a system the company built in early 2026 to sift through the wider Chrome codebase. The AI agent scans historical code changes and previously documented CVEs to identify potential weaknesses, even in legacy features that are no longer under active development.

The AI-Powered Security Pipeline

Google's investment in AI for security is not new. The company has been using machine learning for fuzz testing since at least 2012. However, the integration of large language models has dramatically accelerated the workflow in four critical stages: discovery, triage, patching, and deployment.

In 2023, Google began building LLM-based security systems, followed by a tool called Naptime in 2024 in collaboration with Project Zero. A more advanced agent, Big Sleep, was developed later that year by DeepMind and Project Zero, successfully discovering bugs in the V8 JavaScript engine and the graphics stack. The 2026 Gemini harness added interoperable open-weight and proprietary models, a knowledge base built from Chrome's Git history, and a "critic" agent that processes developer-written SECURITY.md files to understand trust boundaries.

Once bugs are reported, AI now automates much of the triage process. A hybrid pipeline of rules and models filters out spam and duplicates, attempts to reproduce the bug on the affected operating system and Chrome version, attaches metadata such as severity and introduction date, and routes the report to the appropriate engineering team. Google estimates this automation saves developers hundreds of hours every month.

For patching, Google deploys a "fixing agent" that generates multiple candidate code patches, while a second "critic agent" evaluates them to select the strongest solution. The system iterates until the change compiles and meets Chromium style guidelines, even writing tests before an engineer signs off on the final fix. The models operate on locked-down machines with no general internet access, behind an allowlisted network interception, and are restricted from touching files outside designated source directories.

Closing the Patch Gap

Finding and fixing bugs is only half the battle. The "patch gap"—the window between when a fix appears in public source code and when users actually apply the update—remains a critical vulnerability. Attackers can reverse-engineer patches to develop exploits during this period.

To combat this, Google is moving to a two-week release cadence for major Chrome milestones, supplemented by weekly security updates. In the face of what the company describes as "fast-moving, AI-powered attacks," it is now piloting two security releases per week.

"Even with this pace, proper public disclosure remains paramount," Google stated in its report. "Every security bug that reaches Chrome Stable, regardless of whether it was discovered internally or reported externally, is documented and disclosed publicly as a standard best practice."

Google is also tackling the user-experience friction of frequent updates. Chrome typically requires a restart to apply patches, a step many users delay. In Chrome 150, Google introduced a "zero window restart" feature on macOS, where applications often continue running in the background even after all windows are closed. If Chrome detects a pending update while in this windowless state, it now restarts automatically, applying the fix seamlessly.

Looking further ahead, the company is developing "dynamic patching," a technique that leverages Chrome's multi-process architecture to sequentially replace background child processes—such as the Renderer and GPU—with updated binaries on the fly, eliminating the need for a restart altogether.

Structural Defenses and Industry Trends

Beyond patching individual bugs, Google is investing in structural changes to eliminate entire classes of vulnerabilities. The company is hardening its C++ runtime environment with tools like MiraclePtr to combat use-after-free and out-of-bounds memory safety flaws. It is also increasingly rewriting components in Rust, a memory-safe language, and building Chrome's top-level user interface using HTML, CSS, and TypeScript to reduce dependency on traditional C++ frameworks.

All third-party dependencies in Chrome are being moved onto automated update pipelines to ensure they remain current, and AI tools integrated into the continuous integration system blocked more than 20 vulnerabilities from reaching production in May alone, including one critical issue.

The exponential rise in AI-driven vulnerability discovery is not unique to Google. Microsoft patched a record 570 security flaws during its June Patch Tuesday release, also citing internal AI tools as a contributing factor. Apple, however, does not appear to be on the same trajectory. An independent count shows Apple patched 482 bugs in 2026, roughly on pace with its 2025 totals.

Parisa Tabriz, Chrome's vice president and general manager, acknowledged the current spike may represent a transitional phase. "There's this near-term spike, but I do think there's going to be a new equilibrium," Tabriz told Wired. "Across the industry I think it's really important that people who are building and thinking about software security are incorporating AI into their development workflows."

The surge in external bug reports has also forced changes to Google's vulnerability reward program. By March, the company had received more submissions than in all of 2025 combined, prompting a restructuring to steer researchers toward bugs its internal AI tools are less likely to find.

"Every bug found and fixed is one less foothold for an attacker," Google's Chrome Security Team stated. "But discovering and fixing a bug is only half the battle—we must also ship the fix and apply the update for users faster than adversaries can exploit the bug."