At WWDC 2026, Apple unveils a new generation Siri assistant, supported by Google’s Gemini models, capable of acting autonomously in your applications. A milestone for the Californian firm.
After an initial failure, Apple is returning to AI. On the occasion of WWDC 2026, its annual developer conference, Apple presented this Monday, June 8, the new version of its personal assistant Siri. The agent, because yes Siri is now an agent capable of acting directly in the user’s phone to respond to all their requests. A late but welcome alignment with the competition which was at the heart of Apple’s announcements. A look back at all the new features in artificial intelligence that are coming very soon to all Apple devices.
Siri becomes an AI agent
Renamed Siri AI, Apple’s AI is changing its nature. Where the old Siri was content to trigger isolated commands, the new version relies on four bricks which, put together, form a real agent: understanding the personal context (your photos, messages, emails, notes, indexed locally), awareness of what is displayed on the screen, access to the web for up-to-date information, and above all app actions, which allow Siri to dig into the tools of your applications to execute a request from start to finish. butt. All coordinated by a system orchestrator who arbitrates between these capabilities. Apple is careful not to say the word, but the mechanics described are exactly that of an agent. Siri no longer responds, it acts
Concretely, this gives sequences previously impossible on iPhone. In the demos, Siri identifies a location from a photo displayed on the screen, cross-references this information with the address of a contact found in your messages, then generates an itinerary with an intermediate step, without the user opening a single application. This is precisely the promise made by Apple in 2024, which has remained a dead letter until now. We will wait to see it hold up in real conditions before declaring victory, but on paper, the experience is finally close to what ChatGPT, Claude or Gemini already offer.
To materialize this shift, Apple is launching a dedicated Siri application, designed as a conversational chatbot in which you can continue the thread of past exchanges, exactly like its competitors. The assistant also gains a visual dimension (Siri sees what the camera sees and responds live), writing tools integrated wherever you write, capable of adapting to your tone according to your interlocutor, and significantly more precise dictation. Everything is expected to be deployed on all of the brand’s devices, from the iPhone to the Mac including the Vision Pro watch and headset, with on macOS the ability to reason about local files, in RAG mode.
In terms of availability, Siri AI will not be available at launch on iOS and iPadOS in Europe. Apple explains that it is seeking “a path that preserves the confidentiality and security of users”, a formulation which implicitly refers to the regulatory constraints of the DMA and possibly the AI Act. China is also excluded at the start. The rest of the world will be entitled to a beta later this year.
Apple intelligence, a contextual AI system based on Gemini
To function, Siri AI relies exclusively on Apple Intelligence. The system is based on a new generation of Apple foundation model, designed through collaboration with Google and its Gemini models. The models run on the device (probably a model distilled from Gemini) and on the servers via Private Cloud Compute, the in-house infrastructure presented in 2024 and supposed to guarantee that data is neither stored nor accessible, even by Apple.
This new so-called “contextual” intelligence now permeates all native apps. Thanks to Apple Intelligence, Safari sorts tabs by theme and generates extensions based on simple description. The Passwords app can navigate sites alone in the background to replace compromised passwords. The Phone app can report relevant information during a call (example: a confirmation code received by email). And routines in Shortcuts are now created in natural language. Finally, the Photos app inherits a stunning spatial crop, capable of moving the point of view of the shot after the fact, in addition to a generative filling functionality. Certain functions (without Apple specifying which ones) will be capped by daily quotas, with expanded access through iCloud+ subscriptions.
A tempting promise on paper
On paper, Apple’s promise is enticing. The Cupertino company half accepts its delay and claims to have taken the time to think about AI so that it is truly useful to everyone, serving real and personal needs. And in fact, all of the announcements outline what we have long expected from a true personal AI: an assistant who knows your context, acts in your applications and follows you from one device to another. In short, Apple has created a personal agent, undoubtedly less autonomous than an OpenClaw or a Hermes, but designed for your daily life rather than for benchmarks.
This layer of personal intelligence is also accompanied by a reaffirmed commitment to privacy, Apple’s spearhead for years, which positions the brand against the tide of a market largely dominated by players very fond of data. The paradox, all the same, will not escape anyone: to catch up, Apple has entrusted the engine of this AI to Google, precisely the company whose model is based on the exploitation of data.
Finally, there remains the unknown of the calendar and geography, with deployment postponed “into the year.” Apple has often announced before delivering, and the 2024 promise is still awaiting fruition. This time the company perhaps has the means to achieve its ambitions. The demonstration remains to be made.
