Apple iOS CVE-2026-86903

Exploitation requires the ability to execute code on the target device as a normal (non-privileged) app - CVSS AV:L/PR:L/UI:N means a locally installed application can trigger the read without any additional authentication, entitlement or user interaction beyond the app already running. The victim must therefore be induced to install or already have a malicious/hostile app, or the attacker must control a sideloaded/enterprise-signed binary; this is the primary limiting factor, since no remote or unauthenticated path is described. Only devices running builds prior to the version 27 release train are vulnerable; updated systems (iOS 27, iPadOS 27, macOS Golden Gate 27, tvOS 27, visionOS 27, watchOS 27) have the improved input validation in place. The impact ceiling is disclosure of kernel memory only - no memory corruption, code execution, privilege escalation, data modification or denial of service is described, so the primitive must be chained with a separate vulnerability to escalate. No information in the supplied data indicates which specific kernel subsystem or interface is reached, so it is not possible to state whether a sandbox escape, a specific entitlement, or a particular file/format parser is required beyond 'an app'.

CVSS 3.1 scores this 5.5 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N) - a purely local, low-complexity, low-privilege information-disclosure issue with a single-tenant scope. The vector implies the attacker already runs code on the device (PR:L is the app's own process context), so this is not a remote or drive-by bug, and the 'no user interaction' metric simply reflects that the malicious app triggers the read itself. EPSS is only 0.15% (5th percentile), there is no reported exploitation, no CISA KEV entry and no public proof-of-concept in the supplied data; the combination of AV:L and an information-disclosure-only impact keeps real-world urgency moderate. The genuine value of a kernel memory leak is as a primitive: it can defeat kernel ASLR and enable a follow-on privilege-escalation chain, which is why Apple still ships a fix on an emergency-adjacent cadence. Note that the input provides no SSVC decision and no exploit-maturity data; absence of POC in this dataset should not be read as proof that no private exploit exists. Treat as a patch-and-move-on item rather than an emergency, but prioritize on devices where a hostile app model is realistic (BYOD, sideloading-enabled, MDM-managed fleets).

An attacker convinces a user to install a seemingly benign application (or abuses an enterprise/sideloading distribution channel) on an unpatched iPhone, iPad or Mac. When the app passes a crafted, oversized input to the vulnerable kernel interface, the out-of-bounds read returns adjacent kernel memory, which the app exfiltrates or uses locally to defeat kernel address-space randomization. That leak becomes the reconnaissance primitive for a chained kernel privilege-escalation exploit; no public proof-of-concept was identified in the available intelligence.