Google Chrome CVE-2026-106370

Exploitation requires the attacker to have ALREADY compromised the Chrome renderer process via a separate, prior vulnerability; only then can a crafted HTML page trigger the uninitialized GPU resource to read memory outside the sandbox. This is a chained, second-stage sandbox information-disclosure primitive, not a standalone remote attack. Limiting factors: it is not exploitable in isolation (a renderer compromise is a hard prerequisite), it requires user interaction (navigating to attacker-controlled HTML, per UI:R), impact is limited to low-severity memory reads (C:L; no integrity or availability impact), and it is fixed in Chrome 155.0.8059.39.

This is a genuinely low-to-moderate priority issue despite multiple signals, and the raw CVSS of 4.7 (Medium) is an honest reflection. The CVSS vector (AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:N/A:N) shows network reachability with only low confidentiality impact and no integrity or availability impact; the scope change (S:C) reflects a sandbox boundary crossing, which is why it scores above a trivial info leak. However, the vector's PR:N is misleading when read against the description: the description explicitly requires 'a remote attacker who had compromised the renderer process,' meaning this is a second-stage sandbox-read primitive that is NOT exploitable on its own - it must be chained with a separate renderer compromise. EPSS corroborates low urgency at 0.19% (8th percentile), indicating very low predicted exploitation probability. There is no CISA KEV listing and no public exploit code identified, so there is no evidence of active exploitation or weaponization. Chromium itself rates the security severity as Medium. Conflict to flag explicitly: CVSS encodes PR:N (unauthenticated) but the description's precondition of a pre-compromised renderer is effectively a significant prerequisite not captured by the base metrics - defenders should treat this as a sandbox-hardening/defense-in-depth fix, not a standalone remote exploit. Secondary discrepancy: the description states Android while the linked advisory is a desktop stable-channel update (see confidence_notes). Overall: patch via normal Chrome update cadence; not an emergency.

An attacker first compromises the Chrome renderer process using a separate vulnerability, then serves a crafted HTML page that triggers the uninitialized GPU resource to leak memory contents from outside the renderer sandbox, aiding further exploitation or information theft. No public exploit code has been identified for this specific issue, and realistic exploitation depends on chaining it with a prior renderer bug.