Apple Opens Siri AI to All Developers in iOS 27.3: Skills, App Intents, and a Spring 2027 Payoff

Apple on Monday released developer betas of iOS 27.3, iPadOS 27.3, macOS 27.3, and visionOS 27.3, removing its invitation-only waitlist and opening Siri AI integration to any registered developer for the first time in the assistant's 15-year history. The expansion — delivered through the App Intents framework — introduces a new API capability called Skills that lets Siri AI chain actions across multiple apps in a single request, marks the formal end of SiriKit as Apple's Siri developer API, and sets a spring 2027 App Store submission window that will determine when most iPhone users actually see the results.

That timeline gap is the most important thing developers should understand about Monday's announcement. The iOS 27.3 Developer Preview is explicitly a build-and-prepare stage. Siri AI integrations will not appear in the App Store for general users until Apple opens submissions "in spring" — meaning most of the platform's potential value arrives five to six months from now, once apps pass review and reach users through a standard update cycle.

Read more: Apple turns to Alibaba Qwen for China AI after 22 months of regulatory delay

App Intents Replaces SiriKit — and the Replacement Is Architecturally Different

The App Intents framework, first introduced in iOS 16 in 2022 as a Shortcuts companion, is now the only pathway to Siri integration. SiriKit, which Apple formally deprecated at WWDC 2026 in June, will remain supported for two to three more years before removal, but no new SiriKit capabilities are coming, and new Siri AI features are exclusive to App Intents.

The architectural difference between the two frameworks matters for understanding what developers can now build. SiriKit mapped speech to rigid predefined "domains" — a fixed set of categories including messaging, calling, payments, and ride-booking. A developer had to fit their app's capabilities into one of those boxes, which left most app categories with no Siri integration path at all.

App Intents inverts this. Developers define their own intent types — the specific actions their app can perform — and their own entity types: the objects inside the app that Siri AI can reason about. A music app can define a "playlist" entity and an "add track to playlist" intent in its own terms. A task manager can expose projects, due dates, and priorities as entities. Siri AI uses on-device natural language understanding to match what a user says to these developer-defined constructs, rather than forcing the user's request into a predefined category.

Crucially, App Intents definitions are compiled into the app bundle and extracted by iOS at install time. No server-side registration is required, and no Siri-specific review gate preceded Monday's announcement. Any developer who ships an App Intents-enabled app will have that functionality indexed automatically.

Skills: Siri AI Becomes a Cross-App Coordinator

The new Skills API is the most significant capability expansion in Monday's release. Where a standard App Intent gives Siri AI control over a single app action — opening a playlist, adding a task, checking a price — Skills allow Siri AI to combine intents from multiple apps into a coherent multi-step response.

Apple's examples from the Developer Preview documentation illustrate the range: "find a coffee shop I visited last week" requires Siri AI to combine location history, calendar or notes data, and a maps service into a single surfaced result. "Archive files that haven't been opened in over a year" crosses the filesystem with potentially a file-management app's intent.

Developers do not write the cross-app coordination logic themselves. Each app exposes which of its intents are Skills-eligible by marking them with the appropriate API flag; Siri AI reasons about how to chain those contributions to satisfy a user's request. This means the quality of a Skills-based interaction depends both on how clearly individual developers define their intent input and output types, and on Siri AI's reasoning accuracy about how to combine them — a chain that may produce unexpected results during the preview period when the model has not yet been tuned against real developer implementations.

How Siri AI Processes a Request Across Three Tiers

Siri AI routes requests through a three-tier compute architecture, and developers building App Intents integrations should understand where their user's data travels.

The first tier is Apple Foundation Models Core, the on-device language model running on Apple Silicon. Simple queries and intent matching at the app level are resolved here, with data never leaving the device. The second tier is Private Cloud Compute, Apple's cloud AI infrastructure with hardware-rooted trust and no persistent request storage — independent auditors can verify that Apple's privacy claims hold for PCC traffic. The third tier is Google Cloud, where complex queries requiring Gemini-class reasoning are routed, running under Apple's privacy terms on Google infrastructure; the Google partnership, announced in January 2026, underpins this processing tier.

Siri AI routes dynamically across these tiers based on query complexity; developers do not select which tier handles their intent, nor do they receive signals about which tier was used. For Skills chains, different components of a single request may cross tier boundaries, raising questions about data consistency that Apple has not yet fully addressed in the Developer Preview documentation.

Siri AI Reaches Six Languages, With Four More Planned for December

The six languages currently supported — English, French, Japanese, Korean, Portuguese, and Spanish — have been in Siri AI since iOS 27.2, released October 26. Monday's iOS 27.3 Developer Preview adds no new languages; it opens developer access across the existing six-language surface.

Additional languages are planned for the December public iOS 27.3 release: Traditional Chinese, Norwegian, Turkish, and Vietnamese. That December release is also when Apple Intelligence launches in China through Alibaba's Qwen models — though Siri AI itself in China requires separate regulatory approvals from China's Cyberspace Administration and is not part of the December package. China's Apple Intelligence debut with iOS 27.3 covers Writing Tools and other Apple Intelligence features, not the Siri AI agent capabilities that Monday's developer announcement describes.

EU Mobile Users Remain Excluded

Siri AI is unavailable on iOS, iPadOS, and watchOS in the European Union. Apple has not launched the assistant on EU mobile platforms, citing compliance requirements under the Digital Markets Act, which designates Apple as a gatekeeper and imposes interoperability obligations Apple has not yet satisfied for Siri AI specifically. The assistant is available on Mac and Vision Pro in the EU.

The exclusion is significant because the EU represents a substantial portion of iPhone revenue globally. Developers in EU markets building Siri AI App Intents integrations will reach their home users only on Mac and Vision Pro until Apple resolves its DMA compliance position on mobile. No specific EU mobile timeline has been announced.

Apple Warns Investors That Agents May Resist Instructions

Concurrent with Monday's developer beta, Apple filed its FY2026 10-K annual investor report. The filing contains a new risk disclosure stating that autonomous AI agents "may behave in unintended or unpredictable ways" and may also defy directions or resist instructions — language that extends to any third-party technologies Apple deploys.

This language is notable because it represents Apple's first formal acknowledgment in an investor filing that it is building and deploying systems it characterizes as capable of genuine autonomous agency. The disclosure applies to Apple Intelligence broadly, including Siri AI and its Skills-based cross-app capabilities. For developers building Skills integrations, it is a signal that Apple is treating multi-step Siri AI as a genuinely agentic system — with the operational unpredictability that entails during a preview period.

Alexa Opened Years Earlier, But App Intents Is Architecturally Tighter

Apple's platform opening arrives five to six years after Amazon opened the Alexa Skills Kit to any developer, and several years after Google Actions on Google reached general availability. The late timing is genuine, but Apple's integration model differs structurally from both predecessors.

Alexa and Google Assistant skills require server-side hosting: developers build, host, and maintain an endpoint that the voice platform calls at runtime. App Intents requires no server infrastructure — the entire intent definition lives in the app bundle, reducing developer overhead for apps that already ship in the App Store. The same App Intents code that powers Siri AI integration simultaneously powers enhanced Shortcuts workflows and improved Spotlight results, compounding the return on a single implementation investment.

Amazon Alexa Skills Kit has also accumulated years of ecosystem maturity; App Intents is earlier in its trajectory. Google completed its full consolidation of Assistant into Gemini with the September 2026 shutdown, retiring the Actions on Google developer ecosystem and moving the platform onto Gemini's new agent framework.

Apple benefits from a substantially larger active-user base than any competing voice assistant platform in English-language markets, and from tighter hardware-software integration that enables on-device processing unavailable to cloud-dependent alternatives. Whether those structural advantages translate into a more capable Skills ecosystem depends largely on how quickly iOS developers instrument their apps and how accurately Siri AI learns to reason about the resulting intent graph.

What Comes Next for Developers and Users

For developers, the immediate task is implementing App Intents and Skills support during the developer preview period. Apple is running structured Apple Intelligence Group Labs sessions November 4–5 to assist developers with implementation — a signal that the company is treating the preview as an active onboarding window, not merely a documentation release.

The spring App Store submission window is the platform's real commercial inflection point. When Apple opens that gate and apps carrying Siri AI integrations pass review and push updates to users, the size and quality of that first wave will determine whether Siri AI establishes itself as a credible app-integration layer or arrives with too thin a capability set to shift user behavior.

For the EU and China, where the platform is either blocked or restricted on mobile, the trajectory depends on Apple's regulatory work — a timeline that has consistently proven slower than Apple's own product calendar.