Google Chrome EUVD-2026-66010

The victim must actively visit an attacker-controlled or compromised web page containing crafted media content - user interaction is mandatory (UI:R per CVSS). The installed Chrome version must be below 152.0.7977.65 on any supported platform. No authentication is required on the attacker's side (PR:N per CVSS). The non-automatable SSVC rating confirms that bulk automated delivery is not viable - each victim requires individual social engineering or drive-by redirection to the malicious page. The sandbox escape means Chrome's renderer sandbox, normally a strong containment layer, does not limit the post-exploitation impact. There is no indication that a specific Chrome feature flag or non-default configuration must be enabled; the Media component is active by default for any HTML page rendering media.

The NVD CVSS score of 8.8 (High) with AV:N/AC:L/PR:N/UI:R reflects a network-reachable, low-complexity attack requiring no attacker privileges, but demanding one user interaction - visiting a malicious page. However, the NVD vector assigns S:U (Unchanged scope), which is inconsistent with the CVE description explicitly stating execution 'outside the sandbox' - a sandbox escape is definitionally a scope change (S:C) in CVSS 3.1 terms. The true severity is likely higher, closer to 9.6. Chromium's internal 'Medium' severity rating may reflect exploitability difficulty or the need for a chained renderer bug, contrasting with the broad CVSS signal. SSVC confirms Exploitation: none and Automatable: no, meaning no observed in-the-wild use and mass automated delivery is not feasible. EPSS data was not provided. No KEV listing. Despite the absence of active exploitation, the full C:H/I:H/A:H impact combined with sandbox escape justifies urgent patching.

An attacker registers or compromises a website and embeds a crafted HTML page containing specially malformed media content targeting Chrome's Media component. When a victim navigates to the page in a vulnerable Chrome version, the malformed media triggers the input validation flaw, escaping the renderer sandbox and executing arbitrary code in the context of the host OS process - granting full access to the victim's files, credentials, and system. No public proof-of-concept has been identified at time of analysis.