Google Chrome EUVD-2026-35223

The victim must be running a vulnerable Chrome build (any version below 149.0.7827.103) and must actively load attacker-controlled HTML content in the browser - per CVSS UI:R, exploitation cannot occur without that user action, but no authentication, special browser configuration, extension, or non-default flag is required, and the malicious page can be delivered via any normal web channel (link click, iframe, ad, redirect). The main limiting factor is the user-interaction requirement and the assumption the user has not yet auto-updated to 149.0.7827.103; there is no platform restriction noted, so Windows, macOS, and Linux desktop Chrome installations are all in scope.

The CVSS 9.6 vector (AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H) indicates network-reachable, low-complexity, unauthenticated exploitation with full confidentiality, integrity, and availability impact across a security boundary - only user interaction (visiting a page) gates it, which is trivially achievable via phishing, malvertising, or watering-hole attacks. No EPSS score, KEV listing, or SSVC decision was provided in the input, and no public PoC is referenced; absence of these signals is not evidence of safety but does suggest exploitation is not yet widespread. Browser sandbox escapes are historically high-value targets often chained with renderer RCE bugs, so even without confirmed in-the-wild use this should be treated as a high-priority patch for any endpoint running Chrome. One caveat: the description labels this a UI input-validation issue while the impact is sandbox escape, which is an unusual combination - defenders should consult the linked Chromium issue tracker entry (518105731) once it becomes public for full technical detail.

An attacker hosts or injects a crafted HTML page (e.g., via phishing email, malicious ad network, or compromised site) and lures a Chrome user to visit it; when the page is rendered, the malformed input triggers the UI validation flaw and escapes the renderer sandbox, giving the attacker code execution at the browser-process privilege level on the victim's host. From there the attacker can read cross-origin data, persist on the endpoint, or chain to OS-level privilege escalation. No public PoC was identified at time of analysis.