While Microsoft, Google, and NVIDIA race to deploy agentic AI that acts on users' behalf, Apple is deliberately narrowing the scope — betting that usefulness and privacy, not autonomy, are what users actually want.
While the industry chases agentic AI, Apple is betting on usefulness over autonomy — and asking users to trust its privacy pitch, one password at a time.
At this year's Worldwide Developer Conference, Apple made a choice that would have seemed unthinkable a year ago: it barely mentioned agentic AI.
Google I/O, Microsoft Build, and NVIDIA's Computex blowout were wall-to-wall with the term — describing AI systems that can automatically schedule meetings, draft emails, or complete transactions on a user's behalf. Apple, by contrast, spent the bulk of its keynote talking about Siri AI doing things a person actually asked it to do: finding a friend's new address buried in a months-old text thread, adding concert tickets to a calendar, proofreading a document. It responds to commands. It synthesizes data. It doesn't take initiative.
Hardawar made a similar case in his analysis for Engadget: "We'll probably see impressive agentic AI within a decade or so, but I worry about leaping into a world of agents with our current batch of AI models, which can still hallucinate and aren't entirely trustworthy." He called letting current AI agents act entirely on their own "sheer insanity."
Federighi put it plainly at a WWDC press panel: "This is the amount of the Google Assistant we use, which is none."
Apple's public caution about agentic AI is underpinned by a concrete technical stance — one the company has spent years constructing. Its Private Cloud Compute (PCC) system, first introduced to support on-device Apple Intelligence processing, was designed so that data sent to the cloud is "vaporized" immediately after a query is answered. Nothing is stored. Nothing is logged. The architecture is intentional.
But WWDC 2026 introduced a complication: for the first time, some of Apple's AI workloads now run on hardware it doesn't own. The company's new AFM3 Cloud Pro model — which handles "agentic tool use and complex reasoning" — runs on Google's servers, with NVIDIA-powered hardware. It's the first time Apple has trusted third-party infrastructure for a core piece of its AI stack.
The company is up-front about what it's doing and why. "While we absolutely minimize what is sent up to PCC, the critical thing about PCC is, architecturally, that's at that point an efficiency measure," Federighi said at the post-keynote press session, via Ars Technica. "Because PCC itself, by design from the ground up, is going to vaporize any record of that data the moment after it answers your question... This is not stored. It's all in a form where it's completely transient."
Apple is using NVIDIA's Confidential Computing, Intel's Trust Domain Extensions, and Google's Titan security chip to provide layers of protection on the third-party hardware. The company also maintains "a cryptographically verifiable, append-only ledger of all Google Cloud hardware that is part of the PCC fleet," and Apple devices will only trust software signed by Apple on those servers.
Apple's approach to privacy is also a competitive differentiator. "Apple's commitment to privacy with Private Cloud Compute also gives it a leg up on Google and Microsoft — the company says it only uploads relevant data, makes it all anonymous and doesn't track server logs," Hardawar noted.
Apple couldn't resist the allure of agentic AI entirely. And, not surprisingly, it's tied to one of the more provocative new Apple Intelligence features: the ability to automatically change compromised passwords.
The new Passwords app will "securely navigate through websites to sign in and upgrade their accounts to strong passwords," Apple said during the keynote, according to Engadget's WWDC roundup. The feature is opt-in but agentic — the system acts on the user's behalf, on websites, without ongoing supervision.
The risk calculus is obvious to observers. "What else can happen when Apple Intelligence logs into that website? Can I really trust the password it comes up with?" Hardawar wrote. "The goal of making users more secure may be worth the slight loss of agency through Apple's agent, but it's a slippery slope."
This is Apple's narrowest but most concrete bet on agentic autonomy — and it's the feature that most directly tests whether the company's privacy-first framing can survive contact with an AI that actually does things without asking first.
Apple is also using agentic AI for Safari's new "Notify Me" feature, which tracks changes to web pages — price drops, restocks, or updates to a company's terms of service. Hardawar noted it's "certainly far better than keeping a tab open forever and constantly refreshing it," but also observed that the architecture makes it easy to imagine the feature expanding: "Maybe eventually AI could automatically purchase a hot product before it goes out of stock, or send an email to a group of friends if you're eyeing a vacation property's pricing."
Apple is rarely the first to new technology. It didn't make the first MP3 player, smartphone, or ultraportable laptop. It typically enters after others have demonstrated demand, then refines the concept for mass-market usability — and charges a premium for the privilege.
"With Siri AI, Apple has the hindsight of seeing where Microsoft failed with Copilot, and it has a clearer vision of the AI features people may actually want to use," Hardawar observed.
This is a company that watched Google and Microsoft rush AI into every corner of their products — sometimes making those products worse — and chose a different entry point. Whether that restraint is wisdom or tardiness is a question Apple's critics will keep pressing. The first batch of Apple Intelligence features, like notification summaries and Genmojis, were "fairly underwhelming," Hardawar noted, and "in the case of those summaries, led to some embarrassing mistakes."
But the stakes of the approach Apple has chosen are higher than a sluggish launch. The company is arguing that the AI moment doesn't require users to trade privacy for capability — that the two can coexist, if the architecture is built correctly from the ground up. The Passwords app's autonomous password-change feature is the sharp end of that argument: it only works if users trust Apple enough to let it act for them. If that trust breaks — if a single over-helpful line of code locks someone out of their bank account — the entire thesis collapses.
Siri AI and the new Apple Intelligence features will launch as part of iOS 27, iPadOS 27, macOS 27 (dubbed "Golden Gate"), and Apple's other operating system releases this fall. Developer betas are available now; a more stable public beta is expected in July. Siri AI will initially be available in English, with more languages to follow.
The feature set includes: a redesigned Siri interface with Dynamic Island integration on compatible iPhones; a dedicated Siri chatbot with image and document generation capabilities; Visual Intelligence for iPhone cameras and Apple Vision Pro; smart tab organization in Safari; automatic password strength upgrades via the Passwords app; AI-powered writing tools and proofreading across apps; and significant performance improvements — Apple claims AirDrop transfers will be up to 80 percent faster, and apps will launch up to 30 percent faster.
Notably, Siri AI will be delayed in the European Union under iOS 27 and iPadOS 27, due to the Digital Markets Act. It will be available on macOS 27 and visionOS 27 in the EU at launch.
This story was reported and published June 12, 2026 — the day of WWDC 2026 coverage.
Sources: Engadget — "I'm Glad Apple Isn't Hyping Up Agentic AI (Yet)" by Devindra Hardawar; Engadget — "Everything Announced At Apple's WWDC 2026 Keynote" by Kris Holt; Ars Technica — "Apple says its AI is still private, even when it's running on Google's servers" by Andrew Cunningham