Apple Patches Microphone Permission Bypass That Let Sandboxed Apps Record Audio
Vulnerability Disclosure · 7 sources · Updated 5h ago · First seen Nov 16, 2026

Apple patched CVE-2026-65339, a privacy vulnerability that allowed a sandboxed malicious app to record microphone audio even after the user denied permission. Meta Red Team X found that Apple’s audio daemon, audiomxd, trusted an app-supplied audit token, allowing the app to impersonate the daemon and trigger an exemption from Transparency, Consent, and Control (TCC) checks. On an earlier iOS 27 beta build, recording neither activated the orange microphone indicator nor identified the responsible app in system logs. A later demonstration build displayed the indicator, so the lack of a visible warning was build-dependent. Exploitation required installation of a malicious app; it was not a remote attack and did not grant general TCC permissions.
Apple restricted audit-token overrides to callers holding a private entitlement. Fixes included iOS 27.0 beta 5 on August 10 and iOS/iPadOS 26.6.1 and macOS Tahoe 26.6.2 on August 17, with reporting describing additional releases through October. Organizations should verify that managed Apple devices have the applicable updates and maintain controls over untrusted app installation. The finding also reinforces an established XPC security lesson: caller identity must be authenticated rather than inferred from attacker-controlled values. Earlier research into a separate, fixed Malwarebytes vulnerability demonstrated how PID-based validation could likewise misidentify a calling process.


