Microsoft Edge CVE-2026-58596

Exploitation requires the victim to load attacker-controlled web content in Microsoft Edge (Chromium-based) and perform the interaction implied by UI:R (e.g., visiting or clicking through a malicious page) - it is not a silent drive-by against a passive target. No authentication is required (PR:N), so the attacker needs no account or credentials, but AC:H means a working exploit depends on defeating browser memory-safety mitigations and specific memory layout, limiting reliability. The scope change (S:C) indicates the payoff is crossing from the sandboxed renderer into a higher-privilege context. Limiting factors: mandatory user interaction, high exploit complexity, no known public exploit (E:U), and an available official fix that closes the window on unpatched hosts.

Signals are mixed and should temper the raw 8.3 score. The CVSS vector (AV:N/AC:H/PR:N/UI:R/S:C/C:H/I:H/A:H) shows a network, unauthenticated vector with a scope-changing, high-impact outcome, but two limiters stand out: AC:H (a reliable exploit is hard to build, consistent with pointer-dereference bugs that depend on memory layout) and UI:R (a user must open or interact with malicious content). Temporal metrics reinforce a currently-low operational threat: E:U means exploit code is unproven, RL:O confirms an official fix exists, and RC:C means the report is vendor-confirmed. There is no KEV listing and no EPSS score in the input, so probabilistic exploitation likelihood cannot be quantified. Net: a genuinely high-severity memory-safety bug in a very widely deployed browser, but with no evidence of active or public exploitation and non-trivial exploit conditions - patch promptly, but it is not an emergency zero-day.

An attacker hosts a malicious web page (or malvertising / compromised site) crafted to feed an untrusted pointer to a vulnerable Edge code path, then entices a victim to visit it. When the page loads and the victim interacts as required, the dereference is triggered to cross the browser's privilege/sandbox boundary and run attacker-controlled logic at higher privilege. No public exploit is identified at time of analysis, and the high attack complexity means reliable exploitation would require significant engineering.