Google Chrome CVE-2026-17746

Exploitation requires two pre-conditions to be met simultaneously: (1) the attacker must have already compromised the Chrome renderer process on the target machine - this CVE alone does not provide initial access and only serves as a sandbox escape primitive in a chained attack; (2) the target must be running Google Chrome on macOS specifically (Windows and Linux are not listed as affected). User interaction is also required - the victim must visit or have already visited a crafted HTML page that initiated the renderer compromise. The GPU acceleration subsystem must be active (the default). Exploitation is limited by the macOS platform restriction, the requirement for an existing renderer compromise, and the high attack complexity entailed in timing the use-after-free. Systems running Chrome 151.0.7922.72 or later are not vulnerable.

The CVSS 3.1 vector (AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:L = 5.8 Medium) correctly captures key risk dimensions: high attack complexity due to the prerequisite renderer compromise, a changed scope reflecting the sandbox escape, and user interaction for the initial delivery. The EPSS of 0.21% at the 11th percentile signals very limited observed exploitation probability in the wild, consistent with the two-stage exploit requirement - attackers need a separate renderer RCE bug before this becomes useful. There is no CISA KEV listing and no public POC at time of analysis, which aligns with the technical barrier to exploitation. Despite the medium CVSS score, the actual downstream severity of a successful sandbox escape is considerably higher than C:L/I:L/A:L implies - it represents a full sandbox breakout, giving the attacker code execution outside Chrome's isolation boundary. The risk is real but probabilistically low: it requires a sophisticated, chained attack and is mac-specific, limiting the exploitable population.

An attacker who has already leveraged a separate renderer-process RCE vulnerability serves a crafted HTML page to the victim, triggering GPU IPC communication that causes a use-after-free in Chrome's GPU process on the macOS host. The attacker leverages the UAF primitive to achieve controlled memory corruption in the GPU process heap, escaping Chrome's renderer sandbox and executing arbitrary code in the context of the GPU process. No public proof-of-concept exists for this specific CVE, meaning practical exploitation requires significant exploit development capability beyond what is publicly available.