Apple has fundamentally shifted its privacy strategy by integrating Google's Gemini AI model into Siri, marking a dramatic departure from the company's decade-long positioning as the anti-Google alternative. At WWDC 2026 on June 20, Apple unveiled Apple Intelligence 2.0, a rebuilt Siri that can understand personal context, read your screen, and take actions across apps. But the most significant change is one Apple glossed over: Gemini is now one of the engines powering Siri, raising questions about what data flows to Google and how Apple's famous privacy promises hold up when a third-party AI model is in the chain.

For the past decade, Apple's entire brand identity has rested on a simple promise: "What happens on your iPhone stays on your iPhone." The company built App Tracking Transparency, Hide My Email, and Private Relay specifically to keep user data away from Google and other advertising companies. Then came WWDC 2026, where Apple quietly announced that Google, whose entire business model depends on collecting and analyzing user data, is now one of the core engines running Apple's flagship AI feature. Apple has not published detailed specifications about what Gemini can see, what it can access, or what data flows to Google when Siri routes a query through its model.

What Can the New Siri Actually Do?

The rebuilt Siri represents a genuine reset for Apple's assistant, which has long been the subject of jokes about failing to understand basic requests. The new version can understand personal context by searching across your messages, emails, photos, and calendar. It can read what is currently on your screen and answer questions about it. It can take actions across multiple apps rather than simply launching them. Most importantly, it can surface specific details in real-time situations, such as finding an airline confirmation code in Mail while you are on hold with the airline.

Apple has emphasized this last scenario repeatedly, and for good reason. It describes the exact friction point where current Siri forces users to put a call on hold, manually open Mail, search for a confirmation code, switch back to the phone, and read it aloud. The new Siri AI promises to eliminate that entire workflow. Whether it actually delivers on this promise depends partly on whether app developers implement the necessary integration framework, a process that typically takes years rather than months.

How Will Apple Intelligence 2.0 Actually Work for Developers?

The success of Siri AI hinges on a technical framework called App Intents, which allows app developers to connect their content and capabilities to Siri. When an app supports App Intents properly, Siri can perform actions inside that app. When it does not, Siri can only tell you the app exists and launch it for you. This is the unglamorous but critical plumbing that determines whether Apple Intelligence 2.0 lives up to its promises.

Apple's developer ecosystem has several pathways to integrate on-device AI into their applications. The company has released multiple frameworks and tools designed to make this integration straightforward.

• Core AI: Allows developers to integrate on-device AI models into their apps with just a few lines of code, with acceleration across the CPU, GPU, and Neural Engine on Apple silicon devices.
• Foundation Models Framework: Enables developers to build agentic app experiences where the model can pick tools, write steps, and run multi-step tasks automatically.
• App Intents Framework: Connects your app to system experiences like Siri, Shortcuts, Spotlight, and Apple Intelligence, exposing your app's actions and content to these features.

Most users will never hear the term "App Intents," but every time Siri fails to complete a task inside a third-party app, the reason will almost always be that the developer has not implemented App Intents deeply enough. This is the least glamorous part of Apple Intelligence 2.0, but it is probably the most important. A smarter AI model can answer better questions, but an assistant that cannot interact with the apps people actually use is still performing a magic trick with a visible seam.

What Does This Mean for Enterprise and Business Users?

Apple is positioning on-device AI as a critical infrastructure shift for enterprise customers. According to a 2026 study by research firm Omdia commissioned by Apple, a significant portion of organizations are planning to move more AI workloads onto devices rather than relying solely on cloud computing. Among the surveyed enterprise technology leaders, 33 percent of hybrid users are planning to shift more workloads on-device, while 65 percent of existing on-device users are planning to expand their use.

The practical applications are already in production across multiple industries. In retail, on-device AI can count stock accurately in a fraction of the time without requiring a network connection. In healthcare, medical imaging analysis can happen on-device, keeping patient data local. In aviation, on-device systems can synthesize weather, traffic, and operations data into recommended actions. In manufacturing, anomaly detection can catch defects and abnormal vibrations before they cascade into larger problems. In financial services, advisors can get briefed on every client with follow-ups already drafted.

When Will Siri AI Actually Be Available?

Siri AI will launch as a beta later in 2026, not on day one of iOS 27. The beta will be restricted to devices set to English, meaning hundreds of millions of iPhone users in other languages will not have access at launch. The feature will not be available in the European Union on iOS, iPadOS, or watchOS at launch while Apple works through regulatory requirements under the Digital Markets Act. It will also not be available in China while Apple navigates regulatory requirements there.

This means the most significant AI feature in Apple's biggest software release of 2026 will not be available in two of the three largest smartphone markets at launch. The unevenness of AI feature rollouts has been one of the defining frustrations of the Apple Intelligence era, and version 2.0 does not fully resolve that problem.

What Is the Real Test for Apple Intelligence 2.0?

Apple Intelligence 2.0 will ultimately succeed or fail on whether it handles small, ordinary tasks better than users could handle them alone. Finding a flight confirmation code during a call. Writing a follow-up email without it reading like it was written by enterprise software. Editing a photo without opening a separate app. Answering a question about something visible on screen without making the user describe the literally visible context.

These are not dramatic use cases. They are the mundane friction points that accumulate across a day of phone use. The design principle behind Apple Intelligence 2.0 is that AI should make the phone feel like it understands the task better, not like it has added a new category of task on top of existing ones. Whether the integration of Google's Gemini into this system undermines Apple's privacy positioning, or whether the company can maintain meaningful privacy protections even with a third-party model in the chain, remains the central tension of this release.