CVE-2026-121255

A crafted page can expose cross-origin data through Chrome Autofill AI

Google · Chrome Autofill AI for desktop

First documented disclosure · Review checkpoint

Summary

Google reports that a page could obtain data across web origins through Autofill AI.

Severity and exploitation evidence

Medium · numeric CVSS score not recorded. Severity source.

Exploitation not established: Google’s public entry for this CVE does not report malicious exploitation.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The Autofill AI feature applied an incorrect authorization check when a crafted HTML page reached the affected path.

Conditions required

A vulnerable Chrome installation must process the attacker’s page. The public disclosure does not specify further interaction or feature settings.

Potential impact

Data from another origin could be disclosed. The sources do not identify which data fields were exposed.

Affected versions and configurations

Chrome versions before 157.0.8103.38, per the Chrome CNA.

Fix and mitigation status

The desktop stable update ships 157.0.8103.38 on Linux and 157.0.8103.38/.39 on Windows and macOS.

Practical next steps

Deploy a current Chrome build containing this fix and verify the running version after relaunch.

Operating lesson

Test origin checks around autofill suggestions with frames and documents controlled by different sites.

Evidence and limits

Medium is Chromium’s qualitative rating. No numerical CVSS or public technical reproduction was available in the reviewed vendor material.

Sources