Google Chrome CVE-2026-11672

Exploitation requires that the attacker has already compromised the Chrome renderer process on Android - explicitly stated in the CVE description - meaning a separate renderer-RCE primitive must be chained in. The victim must also load attacker-controlled HTML in Chrome for Android at a version below 149.0.7827.103 (UI:R), and the high attack complexity (AC:H) implies the heap overflow likely depends on specific heap state or GPU-command sequencing rather than a trivial single-request trigger. Exploitation is limited to the Android build per the description; desktop Chrome is not named as affected, and post-patch devices on 149.0.7827.103+ are not exploitable via this bug.

The CVSS vector AV:N/AC:H/PR:N/UI:R/S:C/C:H/I:H/A:H captures the realistic profile: network reachable, but high attack complexity, user interaction required (loading a malicious page), and a scope change reflecting the sandbox boundary crossed. This is a chained-exploit primitive - it presumes the attacker already controls the renderer - so it is rarely a standalone risk and is typically paired with a separate renderer RCE bug to form a full browser compromise. There is no EPSS score, no SSVC decision data, and no CISA KEV entry provided, and no public exploit identified at time of analysis, which is consistent with Google's standard practice of withholding bug details until users have updated. Real-world priority is high for Android fleets that browse untrusted content because Chrome auto-updates fast but managed devices can lag; however, absent renderer-RCE pairing in the wild, this CVE alone is unlikely to drive opportunistic mass exploitation.

An attacker lures an Android Chrome user to a malicious page that first triggers a separate renderer-process RCE, then from inside the compromised renderer issues crafted GPU commands that overflow a heap buffer in the GPU process to escape the sandbox and run code at GPU-process privilege. From there the attacker can pivot toward broader device compromise or data theft. No public exploit identified at time of analysis.