Apple releases iOS 27 security fixes as limited Siri AI beta starts

Apple released iOS 27 and iPadOS 27 on 14 September, making a broad set of mobile security fixes available while beginning the English beta rollout of Siri AI. The assistant requires Apple Intelligence-compatible hardware, so its launch reaches fewer devices than the operating-system updates.

The release follows Apple’s June announcement and developer-testing phase. For security teams, the immediate issue is the availability of patched operating systems covering CVE-tracked flaws. Those fixes apply independently of whether an eligible device or user can access the new AI features.

Security fixes span critical system boundaries

Apple’s dedicated security advisory covers numerous operating-system components. The documented impacts include access to sensitive data and restricted files, bypassing privacy preferences, arbitrary code execution, modifying kernel memory and changing network traffic. Apple did not rate the representative entries reviewed for this report, and the available evidence does not support a reliable exact total.

The bulletin does not identify one common root cause. It attributes individual flaws to several implementation classes, including out-of-bounds writes, memory corruption, race conditions, authorisation and permissions failures, path-validation errors, state-management problems and a cryptographic integrity issue.

Among the representative fixes is CVE-2026-84607, a race condition in AVEVideoEncoder. Apple says the vulnerability could let a sandboxed application execute arbitrary code with kernel privileges. The company addressed the issue through improved state management.

A Bluetooth out-of-bounds write, CVE-2026-65414, could allow a remote attacker to cause unexpected application termination or arbitrary code execution, according to Apple. CVE-2026-64752 is a CoreMedia memory-corruption flaw that could lead to arbitrary code execution when processing a maliciously crafted image.

Apple also fixed CVE-2026-84523 in APFS, which could allow an application to terminate the system unexpectedly or write to kernel memory. CVE-2026-84533 concerns a cryptographic issue in Heimdal that could let an attacker in a privileged network position modify network traffic.

Several entries affect controls around accounts, personal information and files. Apple says CVE-2026-20683 could let an application misuse the Sign in with Apple flow to access a user’s Apple Account. CVE-2026-65404 could enable a malicious application to bypass Privacy preferences, while CVE-2026-86878 could expose sensitive user data through a Camera permissions issue.

CVE-2026-84624 addresses a CoreML path-validation flaw. Apple’s advisory says the issue could allow a sandboxed application to access restricted files.

Exploitation and proof-of-concept status remain unknown

The reviewed evidence does not establish that attackers are actively exploiting any of these vulnerabilities. Public proof-of-concept availability is also unknown. Neither status can be inferred from the potential technical consequences listed in Apple’s advisory.

The release materials do not describe a confirmed compromise, unauthorised data access, exfiltration or an identified threat actor. The updates are documented as operating-system security releases, not as a response to a disclosed breach.

The fixed versions are iOS 27 and iPadOS 27. Apple’s advisory does not comprehensively identify every vulnerable predecessor or map each CVE to every compatible device. The existence of a fix therefore should not be read as evidence that every issue affected every earlier release.

Security support extends beyond Siri AI eligibility

Apple’s security advisory lists iOS 27 for iPhone 11 and later. It lists iPadOS 27 for iPad Pro 12.9-inch fourth generation and later, iPad Pro 11-inch second generation and later, iPad Air fourth generation and later, iPad ninth generation and later, and iPad mini sixth generation and later.

Siri AI has a narrower hardware requirement. A compatible device can therefore receive the security fixes without gaining access to the assistant. Its beta status, language support, regional restrictions and usage limitations also constrain the practical reach of the headline feature at launch.

Apple outlines its Siri AI privacy architecture

Apple says Siri AI uses a system orchestrator to access the Spotlight index and App Toolbox through processing performed entirely on the device. For requests handled by Private Cloud Compute, Apple says personal data is not stored or made accessible to the company or anyone else.

Apple also says outside experts can verify the privacy properties of Private Cloud Compute. These are vendor claims rather than independent findings about the implementation released with iOS 27. The reviewed evidence contains no independent technical assessment of Siri AI, the system orchestrator or the Private Cloud Compute protections.

Apple says French, Japanese, Korean, Portuguese and Spanish support will follow the initial English rollout in October 2026. Siri AI is initially unavailable on iOS, iPadOS and watchOS in the European Union. In China, Siri AI and other new Apple Intelligence features remain unavailable while regulatory work continues.

Apple lists English (Australia) among supported Apple Intelligence language variants. That listing does not separately confirm every Siri AI capability or launch-day usage limit for Australian users. The security fixes nevertheless apply to compatible Australian devices independently of access to the AI features.

Deployment is a one-way decision

The documented remedy is installation of iOS 27 or iPadOS 27, and no temporary mitigation is identified in the reviewed material. Organisations can assess the fixed releases through their normal testing and deployment processes while treating Siri AI access as a separate hardware and policy question.

Apple says iOS and iPadOS updates cannot be downgraded to the previous version after installation. Staged deployment is therefore relevant for organisations that need to test application compatibility, confirm device eligibility and assess the practical effects of the beta AI features before expanding the rollout.

Mobile rollout decisions can be assessed more clearly when patch scope, device compatibility and vendor privacy claims are treated as separate questions. Apply the same analytical discipline through The Hack Academy’s online cybersecurity training programme, where listed subjects include cybersecurity fundamentals, networking and cryptography, with self-paced learning and practical exercises alongside theory.