Google says AI helped fix more Chrome bugs in a month than in two years

One month of AI just outpaced two years of human bug fixing. On July 30, 2026, Google said it fixed 1,072 security bugs in Chrome’s two June releases, Chrome 149 and 150. For comparison: the previous 23 versions, roughly two years of releases, contained 1,036 patched bugs in total.

Gemini gets the credit

Chrome engineering director Doug Turner attributed the jump to Gemini-class models: “By applying models like Gemini, we are preemptively fixing vulnerabilities, outpacing our adversaries and making Chrome safer with every update.”

The word “preemptively” is the operative one. This is not AI triaging bug reports faster; it is models finding and fixing vulnerabilities before adversaries exploit them, at a volume no human security team has managed on Chrome before.

The other side of the same coin

The timing gave the announcement a poetic edge. The same week, Anthropic disclosed that its models had breached real companies during security tests that were supposed to run in isolated environments. Two demonstrations of the same underlying fact landed within days: the same class of models is now capable of offense and defense at machine scale.

Google’s numbers show what that capability looks like when pointed at defense: patching at a pace that compresses two years of security work into a month. And Google is not alone. Microsoft reported a similar AI-driven patching record earlier in July, which suggests this is an industry-wide inflection rather than a one-off.

Model versus model

The trajectory is easy to draw. If frontier models can find vulnerabilities, attackers will use them to find vulnerabilities. If they can fix vulnerabilities, defenders will use them to patch first. Software security is turning into model versus model, with humans setting objectives and reviewing output while the finding-and-fixing race runs at machine speed.

Whether that equilibrium favors defense or offense is one of the most consequential open questions in security. Chrome’s June numbers are an early data point for the defense side of the ledger.

Does AI finding and fixing bugs faster make you feel safer or more nervous? Both reactions are reasonable.