Apple iOS CVE-2026-65354

Requires a malicious application to be installed and executed on the target Apple device (iOS/iPadOS or macOS before version 27 / macOS Golden Gate 27); it is a local attack (AV:L) by an already-running low-privileged app (PR:L) that abuses a permissions weakness to escape its sandbox. Limiting factors: the attacker must first get their app onto the device (via App Store distribution, enterprise/sideload provisioning, or social engineering), the CVSS vector indicates user interaction is involved (UI:R), and patched OS versions (iOS/iPadOS 27, macOS Golden Gate 27) fully remediate the flaw. There is no remote or network-based path to trigger this.

This is one of the more serious issues in the Apple wave because it is a genuine sandbox escape, not the confidentiality-only info-disclosure that dominates the rest of the batch. The CVSS 3.1 vector (AV:L/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H, 8.2 HIGH) is internally consistent for a sandbox breakout: local vector, a malicious app already holding low privileges (PR:L), a scope change (S:C) as the app crosses its sandbox boundary, and full C/I/A impact on the escaped-to environment. Counterbalancing signals lower the operational urgency: EPSS is only 0.19% (8th percentile), there is no CISA KEV listing, and no public exploit code - so this is 'no public exploit identified at time of analysis' and NOT active exploitation. It is Apple self-reported with a patch already available. Net: a legitimately high-severity technical impact but low near-term exploitation probability, so it ranks as a real priority to patch (sandbox escape is a foundational security-boundary break, often chained after a malicious app is installed) without evidence of widespread or targeted exploitation. One caveat to verify: vendor set UI:R, yet a malicious app typically executes on its own once run - the user-interaction requirement may overstate a gating factor.

An attacker publishes a seemingly benign app that passes App Store / notarization review or is sideloaded, and once the victim runs it the app exploits the permissions flaw to break out of its sandbox and gain access to resources and data outside its confinement. No public proof-of-concept exists, and the local attack vector means the attacker must first get their app onto the device.