Google Chrome CVE-2026-106403

Exploitation requires that the attacker has ALREADY compromised a Chrome renderer process through a separate prior vulnerability - this is the primary limiting factor and makes the bug a chained, second-stage flaw rather than standalone remote compromise. The victim must then load attacker-controlled crafted HTML (UI:R - user interaction via navigating to a malicious or compromised page). The specific component abused is Chrome's Accessibility feature, exploited via an incorrect-authorization flaw (CWE-863) to defeat site isolation. Affected: Google Chrome prior to 155.0.8059.39; fixed in 155.0.8059.39 and later, so patched installations are not exploitable.

Signals are mixed and point to a moderate, chained risk rather than a frontline emergency. The input CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:R, score 8.1) suggests an easy remote unauthenticated attack, but this CONFLICTS with the description, which explicitly requires 'a remote attacker who had compromised the renderer process' - i.e., a prior sandbox/renderer compromise is a hard prerequisite, so this is a second-stage site-isolation bypass, not an initial-access bug. Chromium itself rated the severity only Medium despite the 8.1 base score, consistent with that precondition. The CISA SSVC framework reinforces lower urgency: Exploitation=none and Automatable=no, though Technical Impact=total (full site-isolation bypass enabling cross-origin data theft). EPSS is very low at 0.18% (7th percentile), there is NO CISA KEV listing and NO public POC identified, so there is no public exploit identified at time of analysis. Net assessment: real but conditional risk - meaningful as part of an exploit chain against Chrome, but low near-term widespread-exploitation probability. Patch promptly given Chrome's ubiquity, but this is not a drop-everything KEV-grade item.

An attacker who has already achieved code execution inside a Chrome renderer process (via a separate prior exploit) lures the victim to a page containing crafted HTML that abuses the Accessibility component's incorrect authorization to defeat site isolation. The compromised renderer then reads sensitive data belonging to other origins/sites that site isolation was meant to keep separated. No public exploit code was identified at time of analysis.