Apple just built the most powerful personal assistant ever—and then told 450 million Europeans they can't have it because their own regulators won't let them keep it safe.
The Summary
- Apple announced Siri AI for iOS 27, featuring smarter, more conversational capabilities—but it won't launch in the EU due to a standoff with regulators over the Digital Markets Act.
- The EU's DMA requirements would force Apple to give any AI system "nearly unlimited access" to user devices, including the ability to read messages, make purchases, and execute actions across apps autonomously.
- Apple proposed a compromise solution called Trusted System Agent—an intermediary layer that would let third-party assistants access Siri-level capabilities safely—but the European Commission rejected it outright.
- No timeline exists for when EU users will get Siri AI, marking the first major AI feature gap between US and European iPhone users.
The Signal
This is what happens when regulation designed to limit Big Tech gatekeeping collides with the reality of agentic AI. Apple's Siri AI represents the company's first real move toward autonomous agents—assistants that can act across apps, make decisions, and execute tasks without constant human approval. The EU's Digital Markets Act was written to prevent platform lock-in. But its requirements for interoperability now mean Apple must give any third-party AI the same deep system access it gives its own assistant.
Apple's concern is specific: AI systems can be hijacked to steal passwords and photos, alter files, and change account settings without user consent. Security researchers have already demonstrated these attacks. Unlike traditional apps—which request discrete permissions for specific actions—agentic AI needs broad, continuous access to be useful. An assistant that can "book me a flight to Berlin" needs to read your calendar, access your email for confirmation codes, use your payment method, and modify your travel apps. That level of access, multiplied across dozens of potential third-party AI assistants, is a different security model than iOS has ever shipped.
"According to EU regulators, the DMA requires Apple to give any AI system nearly unlimited access to a user's device, as well as the ability to act on that access autonomously without a user's ongoing visibility and control."
Here's what makes this fascinating: Apple didn't just refuse. They built Trusted System Agent, a proposed intermediary that would let any virtual assistant—Alexa, Google Assistant, whatever comes next—access the same capabilities as Siri AI while maintaining security boundaries. They offered an 18-month gradual rollout. The European Commission said no to all of it. Not "revise this proposal." Not "meet these additional requirements." Just no.
That tells you two things:
- The EU believes forcing open access is more important than Apple's security model, even for autonomous agents
- Brussels hasn't internalized what "autonomous" means in 2026—agents that take actions without asking permission every time
- Apple thinks this fight is worth delaying a flagship feature in a 450-million-person market
The US-EU AI split is now real. American iPhone users will get an assistant that can compose emails, manage their calendar, handle purchases, and coordinate across apps. Europeans won't. Not because of technical limitations, but because their regulators and Apple fundamentally disagree on whether interoperability trumps security when the thing you're opening up can act on its own.
The Implication
Watch what other AI companies do next. If Google, Microsoft, or OpenAI want to ship agentic assistants in the EU, they face the same DMA requirements. Either they comply and build systems with the broad access the Commission demands, or they join Apple in delaying features. The first company to ship a compromised agent that gets exploited in the EU will define this debate permanently.
For builders: the security model for autonomous agents is unsolved. Apple's Trusted System Agent proposal is the first serious attempt at a framework that preserves both interoperability and safety. If the EU won't accept it, someone needs to propose something better. Because agents are coming, and permission dialogs won't scale when your assistant is making 50 decisions an hour on your behalf.