Google Chrome CVE-2026-8568

Requires pre-existing compromise of Chrome renderer process through a separate vulnerability or exploit chain - this is NOT a standalone initial access vulnerability but a post-exploitation technique. User must interact with attacker-controlled content (visit malicious HTML page) per CVSS UI:R requirement. Chrome version must be prior to 148.0.7778.168 and have AI features enabled (default configuration in affected versions). The high attack complexity (AC:H) indicates additional conditions likely apply, such as specific timing windows, race conditions in AI policy evaluation, or particular sequences of API calls that trigger the policy enforcement gap. No authentication is required (PR:N), meaning any remote attacker who can deliver content to the user can attempt exploitation once renderer compromise is achieved. The limiting factors are the prerequisite renderer exploit and user interaction, making this a second-stage attack rather than a direct threat to patched systems or users practicing careful browsing habits.

Real-world risk is moderate-to-low despite the critical nature of Site Isolation. Multiple factors converge to limit practical exploitability: First, the attack requires pre-existing renderer compromise, which itself demands a separate exploitation chain (CVE vector AC:H reflects this). Second, EPSS score of 0.02% (4th percentile) indicates security researchers assess exploitation probability as very low in practice. Third, the vulnerability is not listed in CISA KEV, confirming no detected active exploitation. Fourth, user interaction is required (UI:R), adding social engineering friction. The CVSS score of 3.1 (Low severity) aligns with the EPSS assessment - this is correctly rated for its multi-stage nature. However, the confidentiality impact, though rated Low, involves breaking a fundamental browser security boundary. For organizations already at Chrome 148.0.7778.168 or later, risk is fully mitigated. For those on earlier versions, the vulnerability represents a defense-in-depth concern rather than an immediate crisis: an attacker who has already compromised the renderer can escalate to cross-site data access, but initial compromise remains the primary security challenge.

An attacker first delivers a malicious webpage containing an exploit for a separate Chrome renderer vulnerability (e.g., memory corruption in JavaScript engine). Once the renderer process is compromised, the attacker leverages the AI policy enforcement weakness by crafting specific HTML structures that trigger cross-origin data access requests. The AI component's insufficient policy checks allow the compromised renderer to read sensitive data from other sites the user has open in different tabs (cookies, form data, session tokens) that should be isolated. User interaction is required to visit the initial malicious page, and attack complexity is high due to the multi-stage exploitation chain and need for precise timing or conditions in the AI policy enforcement code paths.