Google Chrome EUVD-2026-35219

Exploitation requires that the attacker has already compromised the renderer process of Chrome prior to 149.0.7827.103, meaning a separate renderer RCE (e.g., a V8 or Blink bug) must be chained first; this CVE alone is not a standalone entry point. The user must also visit and load attacker-controlled content (CVSS UI:R), and the Plugins code path must be reachable in the victim's browsing session. Limiting factors: the prior-renderer-compromise prerequisite substantially raises the bar versus the PR:N CVSS implies, the scope is unchanged (S:U) so no sandbox escape to the host OS is granted, and availability impact is none (A:N) - the practical payoff is cross-site data theft or impersonation within the browser, not host takeover.

The CVSS vector AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:N produces an 8.1 score driven by network reach, low complexity, and high confidentiality and integrity impact, but the realistic risk is materially lower because the description explicitly requires an attacker to have already compromised the renderer process - a precondition not modeled cleanly in CVSS PR:N. EPSS at 0.02% (4th percentile) and the absence of CISA KEV listing or any public PoC are consistent with this being a sandbox-hardening fix typically chained with a separate renderer RCE rather than a standalone exploit. The Chromium-assigned severity of High reflects defense-in-depth importance, not standalone exploitability. A meaningful caveat: the CVSS PR:N rating conflicts with the description's stated prerequisite of a compromised renderer, so PR:N should be read as 'no Chrome login required' rather than 'no prior foothold required' - defenders should treat this as a chain component, not an entry point.

An attacker first exploits a separate renderer-process vulnerability (for example, a V8 type confusion delivered via a malicious ad or compromised website) to achieve code execution inside the sandboxed Chrome renderer. They then load a crafted HTML page that triggers the Plugins origin-validation flaw to reach into or act on behalf of a different origin loaded in the same browser - for example, reading authenticated session data from a banking or webmail tab - without escaping the renderer sandbox entirely. No public exploit identified at time of analysis.