On 8 June 2026, Apple unveiled the largest Siri overhaul in the assistant’s 15-year history. A 1.2-trillion-parameter AI system powered by Google Gemini on Nvidia Blackwell B200 GPUs, spanning on-device processing, Private Cloud Compute, and confidential cloud inference. Moments later came the confirmation: 450 million iPhone and iPad users across the European Union would not get it.
The Digital Markets Act, designed to make digital markets fairer and more contestable, had collided with a product whose system-level integration makes the compliance question hard. This is an architectural standoff with no obvious resolution, and it is nothing like browser choice screens or app store payment links.
This pillar maps the four dimensions of the story — what Apple actually built, why it is blocked, how the platform-by-platform split works, and what the lockout means for the competitive landscape and your own AI strategy. Each dimension gets a full, standalone treatment in its own article in this series.
In This Series
Siri AI and the Google Gemini Deal Powering Apple’s Biggest Assistant Overhaul — The product architecture: how the three-tier privacy model works, why Apple licensed Google Gemini, and the build-vs-buy trade-offs.
Why Siri AI Is Blocked on EU iPhones and Who Bears Responsibility — The regulatory standoff: DMA Article 6(7), the rejected Trusted System Agent, and the two-sided responsibility analysis.
The Siri AI Platform Split in Europe and What It Costs Apple — The consequences: why Vision Pro gets Siri AI in the EU while iPhone does not, and the financial and precedent stakes.
Apple’s Anti-Companion AI Strategy and What the EU Lockout Changes — The strategic implications: Craig Federighi’s “not a companion AI” positioning, the Google competitive comparison, and what EU-based teams should consider.
What is Siri AI and what makes it different from the Siri you use today?
Siri AI is a rebuilt assistant running on a 1.2-trillion-parameter custom Google Gemini model. It uses a three-tier architecture: on-device processing for simple queries, Apple’s Private Cloud Compute for moderate-complexity tasks, and Google Cloud with Nvidia confidential computing for the heaviest reasoning. It understands screen context, performs multi-step reasoning across apps, and maintains personal-context memory across sessions.
The version of Siri on your iPhone today is a 2010s-era product. It pattern-matches commands against a fixed intent library with no reasoning and no cross-app awareness. Siri AI replaces that architecture.
The three-tier processing model is the real story. Each tier determines what data leaves your device and under what trust guarantees. Tier 1 handles roughly 60 to 70 percent of queries on-device with nothing leaving the phone. Tier 2 processes on Apple Silicon servers with cryptographic attestation and no data retention. Tier 3 runs inside Nvidia hardware-enforced trusted execution environments that even Google cannot inspect. Tokenisation and anonymisation happen before any data moves upward.
Legacy Siri was the subject of a $250 million class action settlement over accidental recordings from 2011 to 2019 where conversations were captured and reviewed by human contractors. Siri AI is a clean-sheet rebuild designed to make that failure mode impossible.
For the architecture walkthrough, including how Apple distilled its own Foundation Models for on-device use and why confidential computing makes the cloud tier defensible, see the full architecture breakdown.
Why is Siri AI being withheld from iPhones and iPads in the European Union?
Siri AI is not technically banned. The European Commission has not issued a prohibition order. What exists is a regulatory stalemate: Apple cannot launch Siri AI in the EU without complying with the Digital Markets Act’s interoperability obligations, Apple argues that DMA compliance for an AI assistant with system-level access is not possible at current security standards, and the Commission has rejected every proposed compromise without offering an alternative.
The result is a block affecting roughly 150 million EU iPhone users.
The DMA became applicable in March 2024 and designated six gatekeepers, Apple among them. For iOS, Article 6(7) is the specific obligation: gatekeepers must provide third-party providers with “effective interoperability with the same hardware and software features” the gatekeeper’s own services access. Because Siri AI reads your screen, accesses messages, and performs cross-app actions, Apple argues that granting equivalent access to any competing AI assistant creates a security surface that cannot be managed.
The Commission’s counter, articulated by spokesperson Thomas Regnier, is that nothing in the DMA prevents Apple from launching Siri AI. The company simply must provide the same access to competitors. The disagreement centres on what compliance means in the context of a deeply integrated AI assistant.
For the detailed analysis of who is responsible, including the DMA mechanics, Apple’s attempted compromises, and the evidence on both sides, see the full regulatory analysis.
What is the Digital Markets Act and why does it apply to Siri AI on iOS?
The Digital Markets Act is an EU regulation that designates large digital platforms as gatekeepers and imposes obligations designed to make digital markets more contestable. It applies to Siri AI because iOS is a designated core platform service, not because of Apple’s market share in AI but because Apple controls the platform on which AI assistants run.
The DMA’s quantitative thresholds (€7.5 billion EU turnover, 45 million monthly active EU users) are what iOS satisfies and platforms like visionOS do not.
The DMA’s logic is structural. It targets platforms that serve as important gateways between business users and consumers, regardless of whether the platform owner has done anything wrong. iOS was designated in September 2023 alongside the App Store and Safari, triggering obligations across anti-self-preferencing, data portability, anti-steering, and interoperability. The Siri AI dispute centres on Article 6(7), but it sits within a broader compliance landscape where Apple is simultaneously negotiating connected-device interoperability, browser data portability, and alternative payment provisions.
Understanding why iOS is designated while visionOS is not — and what that means for Siri AI availability across Apple’s product line — requires understanding the DMA’s gatekeeper designation process itself. The quantitative thresholds are the entry point, but the Commission also considers qualitative factors like market entrenchment. For the full analysis of the gatekeeper designation mechanics and what they mean for Siri AI availability across Apple’s product line, see the standoff breakdown.
What does DMA Article 6(7) actually require Apple to do for competing AI assistants?
Article 6(7) requires Apple to provide third-party AI assistants with access to the same operating system features, hardware capabilities, and software functions that Siri AI itself uses, at no cost and on equivalent terms. If Siri AI can read your screen to answer “what’s the address in that text message?”, a competing assistant must be able to do the same through the same OS-level mechanisms, not through a reduced-capability proxy interface.
The obligation extends to microphone access, app intents, notification content, and cross-app automation hooks. The phrase “effective interoperability with the same hardware and software features” is the operational term of art and the centre of the dispute. Apple’s argument: Siri AI has system-level access that no third-party application should receive because granting it creates an attack surface Apple cannot secure. The Commission’s argument: if Apple’s own assistant can access these features, withholding equivalent access from competitors is the type of self-preferencing the DMA was designed to prevent.
The practical mechanism is the interoperability request portal. As of March 2026, 56 formal requests had been submitted to Apple under Article 6(7). The Free Software Foundation Europe reported that none had resulted in a working solution. The detailed analysis of who is responsible examines what reasonable compliance could look like and whether Article 6(7) as drafted is compatible with secure AI assistant operation.
What is Apple’s Trusted System Agent proposal and why did the European Commission reject it?
The Trusted System Agent was Apple’s proposed intermediary software layer — an API mediation interface that would give competing AI assistants programmatic access to iOS features (microphone, screen context, app intents) through a standardised, Apple-controlled interface rather than direct OS-level access. The Commission rejected it as insufficiently “effective” interoperability: because Apple controls the mediation layer, it creates a persistent structural advantage for Siri AI, which accesses those same features directly without going through the intermediary.
The Trusted System Agent represents Apple’s attempt to find a middle ground — comply with the interoperability obligation while preserving a security boundary. In Apple’s framing, the TSA would extend Private Cloud Compute-level protections to competing assistants: queries would be anonymised, data would be ephemeral, and the mediation layer would enforce the same privacy guarantees Siri AI’s architecture provides. The Commission’s counterargument is that “effective” interoperability means same mechanisms, same access — not a proxy that Apple controls and could degrade or throttle.
Apple paired the TSA proposal with an 18-month exemption request, asking for phased compliance rather than immediate interoperability at Siri AI’s launch. The Commission rejected both together, signalling that it views AI assistant interoperability as urgent and that the TSA was a form of compliance that defeats the regulation’s purpose. The rejection left no negotiated alternative on the table, which is why Apple engineers reportedly stopped working toward an EU iOS solution. For the full analysis of the TSA’s technical details and the legal reasoning behind the Commission’s rejection, see the standoff analysis.
Why is Siri AI available on Mac and Vision Pro in the EU but not on iPhone and iPad?
The platform split is a direct consequence of how the DMA’s gatekeeper designation works: it targets specific core platform services, not companies as a whole. iOS and iPadOS are designated core platform services, triggering full Article 6(7) interoperability obligations. visionOS is not designated because Vision Pro’s EU install base (roughly 15,000 to 25,000 units) falls well below the DMA’s quantitative thresholds. macOS is a more complex case: while designated as an operating system, the Commission’s enforcement focus for virtual assistant interoperability has centred on iOS, and Siri AI on macOS 27 will be available in the EU at launch without restriction.
The result is that roughly 0.006 percent of EU Apple users who own a Vision Pro get Siri AI, while 150 million iPhone users do not. This is not a loophole Apple exploited. It is a structural feature of a regulation designed to target dominant platforms, not nascent ones.
The platform split also extends to watchOS 27, which receives Siri AI in the EU alongside visionOS and macOS. But since Siri AI on watchOS requires a paired iPhone with Siri AI, and EU iPhones will not have it, the Apple Watch feature is effectively unavailable too. For the platform-by-platform breakdown, including the financial stakes and fine exposure, see the platform-by-platform breakdown.
What Siri AI features do EU iPhone and iPad users miss, and which ones do they still get?
EU iPhone and iPad users on iOS 27 keep legacy Siri — pattern-matched commands with no contextual awareness, no multi-step reasoning, and no personal memory. They miss Siri AI’s dedicated app with typing-first interaction, screen-aware contextual responses (the ability to answer questions about what is on your display), Camera Siri mode for real-world visual queries, cross-app actions that chain across multiple applications, and persistent personal context that remembers preferences and relationships across sessions. Apple Intelligence features — writing tools, image generation, notification summarisation — remain available on supported EU devices.
The feature gap is the concrete dimension of a dispute that is otherwise fought in legal filings and press statements. Legacy Siri on iOS 27 is the same assistant architecture that predates the DMA — it cannot read your screen, it cannot reason across multiple apps, and it cannot remember what you asked it five minutes ago. Siri AI, by contrast, is designed to function as a system-level orchestrator: “find the address in my last text from Sarah and add it to my calendar” is the type of query legacy Siri cannot parse and Siri AI handles natively. The dedicated Siri AI app, which replaces the pop-up overlay with a full-screen typing and voice interface, is also absent from EU iPhones.
Apple Intelligence features — the broader AI suite released in 2024 — are available in the EU after a compliance negotiation resolved their own separate DMA concerns. Writing tools, Genmoji, Image Playground, and notification summaries all shipped on EU iPhones with iOS 18.4 in April 2025. The distinction matters because news coverage often conflates Apple Intelligence and Siri AI: the former is an EU-available feature suite; the latter is the specific assistant that is blocked. Understanding exactly what you lose — and what you keep — is essential for deciding whether the lockout affects your daily use or your product plans. For the full feature-by-feature breakdown, see the consequences and precedent analysis.
Who is actually responsible for this lockout — Apple, the European Commission, or both?
Responsibility is genuinely distributed. Apple’s side: Article 6(7) as drafted makes secure AI assistant operation structurally difficult because it requires granting unmediated OS-level access to any third-party assistant that requests it, and Apple’s three-tier privacy architecture cannot guarantee user data protection under those conditions. The Commission’s side: Apple had years to design iOS for DMA compliance (the regulation passed in 2022, gatekeeper designation came in 2023) and chose an architecture that creates the conflict; the Trusted System Agent was a form-over-substance compliance attempt. The impasse resists a clean answer.
The evidence cuts both ways. Apple’s €500 million DMA fine in April 2025 for App Store anti-steering violations shows the Commission enforces the regulation seriously. The Commission’s March 2025 connected-devices interoperability specification, which Apple complied with, shows Apple can and does comply when the technical requirements are bounded. But Siri AI’s system-level integration is different from payments, browser choices, or NFC access. No previous DMA compliance action has required a gatekeeper to give third parties the equivalent of an AI assistant’s full system access.
For the full two-sided analysis with the evidence for and against each position, see the responsibility deep dive.
How does the Siri AI DMA dispute compare to Apple’s previous EU antitrust battles?
The Siri AI lockout is the latest and most consequential chapter in an escalating EU-Apple regulatory conflict. The Apple Intelligence delay in 2024 followed the same pattern — EU launch withheld pending DMA compliance — but resolved within 10 months through negotiation because the feature set was discrete. The €500 million App Store anti-steering fine in April 2025 demonstrated the Commission’s willingness to penalise. The July 2026 EU General Court ruling upholding Apple’s gatekeeper designation closed Apple’s procedural escape route. Each dispute has escalated further and taken longer to resolve than the last.
The pattern is worth understanding because it predicts how the Siri AI dispute will likely unfold. The 2024 Apple Intelligence delay was resolved within 10 months: Apple negotiated compliance measures for specific features (writing tools, image generation, notification summaries) and launched them in the EU via iOS 18.4 in April 2025. The App Store anti-steering fine followed the same month — the Commission’s first DMA penalty — signalling that negotiation alone would not suffice. The July 2026 General Court ruling was the latest escalation: Apple challenged its gatekeeper designation on the grounds that iOS does not constitute an important gateway, and the court dismissed all three challenges, establishing a sequencing rule that prevents gatekeepers from pre-emptively litigating DMA obligations before a specific enforcement decision.
Siri AI differs from all three precedents in one critical respect: it is not a discrete feature set that can be negotiated incrementally. It is a deeply integrated system-level assistant whose interoperability demands are total, not feature-by-feature. This structural difference explains why the dispute has no resolution timeline. For the full precedent analysis and the July 2026 court ruling details, see the fine and precedent breakdown.
Can DMA-mandated interoperability and Apple’s on-device privacy model actually coexist?
The tension is structural, not tactical. Apple’s three-tier privacy architecture depends on controlled trust boundaries: data moves from device to Private Cloud Compute to Google Cloud only through gated, anonymised pathways with hardware-enforced guarantees at each boundary. DMA Article 6(7) requires that competing assistants access the same capabilities through the same mechanisms — which means they must operate inside those trust boundaries, or Apple must dismantle the boundaries entirely. The question is whether a mediated access model can satisfy “effective interoperability” without breaking the privacy guarantees that make the architecture defensible.
Apple’s privacy model is the company’s most valuable brand asset and an architectural commitment: stateless compute, cryptographic attestation, hardware-enforced trusted execution environments, and data tokenisation before any query leaves the device. These are observable properties of the three-tier architecture. The DMA’s interoperability obligation is equally real: competition law that targets the structural advantages of platform control, and it applies symmetrically to all six designated gatekeepers.
The question is whether the regulation’s framers anticipated the case where a gatekeeper’s security architecture and the interoperability obligation are in tension, not as a pretext but as an engineering reality. The Commission’s view is that Apple could design for both compliance and privacy. Apple’s view is that the Commission is asking for something not possible under current constraints. For the strategic implications of this tension, see the strategic analysis.
How does the EU lockout affect the competitive landscape between Apple and Google in AI?
The lockout creates an asymmetric competitive landscape: Google’s Gemini is available on Android in the EU without regulatory constraint, while Apple’s Siri AI is blocked on the platform where it matters most. Apple’s flagship AI feature runs on Google’s Gemini model under a $1 billion-per-year licensing deal, while Google’s competing Android assistant runs on Google’s own models and faces no EU availability restriction. For EU consumers, Android is currently the only mobile platform with a modern AI assistant, giving Google a temporary but significant advantage in the European market.
The competitive asymmetry goes deeper than availability. Apple’s flagship AI feature runs on Google’s Gemini model under a $1 billion-per-year licensing deal, while Google’s competing Android assistant runs on Google’s own models and faces no EU availability restriction. Apple is paying its largest mobile competitor for the model that makes its assistant competitive, and that same competitor ships its own assistant to the EU market without restriction.
For EU consumers, Android is the only mobile platform with a modern AI assistant. For developers building AI-powered features, Android is the only mobile platform where EU users can access one. This tilts the developer ecosystem toward Android in Europe, compounding the competitive effect beyond consumer choice.
For the full competitive analysis, including Craig Federighi’s public framing of Siri AI as “not a companion AI” and what the lockout means for Apple’s ability to compete in the EU market, see the competitive and strategic breakdown.
What should you consider when planning an AI product strategy given the Siri AI uncertainty in Europe?
The primary considerations are platform availability (Siri AI is not available on EU iPhones, with no resolution timeline), competitive asymmetry (Android’s Gemini is available in the EU), and regulatory dependency risk (integrations that depend on gatekeeper-controlled AI capabilities can become collateral in future DMA disputes). If your user base is meaningfully EU-based, an AI assistant strategy built around Siri AI is not viable today. The relevant framework is “build for platform-agnostic AI delivery, or accept that your EU users access a different product.”
The core variables you will weigh: user geography (what percentage of your users are in the EU?), competitive pressure (are your competitors shipping AI features to non-EU users that you cannot match in Europe?), integration depth (is Siri AI integration core to your product architecture or a nice-to-have?), and regulatory trajectory (is the DMA dispute likely to resolve, escalate, or stall?). These are not questions with fixed answers; they are dimensions you weight against your own product priorities.
For EU-based development teams, the lockout creates a practical friction: you cannot test Siri AI integrations on EU iPhones because Siri AI does not run on EU iPhones. This means testing on non-EU devices or using simulator environments that do not replicate real Siri AI behaviour. For the full decision framework, see the CTO strategy framework.
Resource Hub: Siri AI and the DMA Lockout — Deep Dives
The Product and the Regulatory Standoff
Siri AI and the Google Gemini Deal Powering Apple’s Biggest Assistant Overhaul walks through the three-tier privacy architecture: on-device processing, Private Cloud Compute, and Google Cloud with Nvidia confidential computing. Covers the $1 billion Google Gemini licensing deal, the build-vs-buy trade-offs, and why Apple’s architecture makes the DMA interoperability question hard. Best read first if you want to understand what was actually built before examining why it is blocked.
Why Siri AI Is Blocked on EU iPhones and Who Bears Responsibility analyses the regulatory standoff from both sides: DMA Article 6(7) obligations, the rejected Trusted System Agent proposal, the denied 18-month exemption, and the evidence for and against Apple’s and the Commission’s positions. Best read second, building on the product knowledge from the architecture article.
The Consequences
The Siri AI Platform Split in Europe and What It Costs Apple explains why Vision Pro and Mac users in the EU get Siri AI while iPhone and iPad users do not. Quantifies the financial stakes (DMA fines up to 10 percent of global annual turnover), maps the precedent pattern across Apple’s EU disputes, and examines the July 2026 court ruling that closed Apple’s procedural escape route. Best read third, extending the standoff into its most counterintuitive consequence.
The Strategic Implications
Apple’s Anti-Companion AI Strategy and What the EU Lockout Changes analyses Craig Federighi’s “not a companion AI” positioning as strategy, the Google Gemini-on-Android competitive comparison, and offers a decision framework for teams evaluating AI assistant strategy during the EU lockout. Covers build-vs-wait criteria, platform dependency risk, and DMA compliance awareness for product teams. Best read last, synthesising the product, regulatory, and competitive dimensions into an actionable framework.
Suggested reading order: Start with the product architecture article to understand what Siri AI is and how it works. Move to the regulatory standoff to understand why it is blocked. Continue to the platform split for the consequences and precedent. Finish with the strategy synthesis for the competitive and decision-making implications.
Frequently Asked Questions
What happened when Apple challenged its DMA gatekeeper designation in EU court in July 2026?
The EU General Court dismissed all three of Apple’s challenges on 8 July 2026, upholding the Commission’s designation of iOS, the App Store, and Safari as core platform services. The court established a sequencing rule: gatekeepers cannot pre-emptively challenge DMA obligations before a specific enforcement decision is issued. Apple retains the right to appeal to the Court of Justice of the European Union on points of law only, a process that typically takes 18 to 24 months. The ruling closed Apple’s procedural path to escaping DMA obligations on iOS. For the full precedent analysis, see the consequences and precedent article.
What are the potential DMA fines Apple faces if it never brings Siri AI to the EU?
The DMA permits fines of up to 10 percent of global annual turnover for non-compliance — approximately €38 billion at Apple’s current revenue. In practice, the Commission has not imposed maximum fines: the €500 million App Store anti-steering penalty was roughly 0.13 percent of Apple’s turnover. A realistic range for sustained Siri AI non-compliance is €1 billion to €5 billion, potentially escalating with periodic penalty payments of up to 5 percent of average daily turnover if the Commission treats it as ongoing non-compliance. See the full consequences picture for the full quantification.
How does the UK’s DMCC differ from the EU’s DMA, and why are UK users also locked out?
The UK’s Digital Markets, Competition and Consumers Act is a parallel but structurally distinct regulatory regime — it shares the DMA’s goals of platform contestability but operates through the UK’s Competition and Markets Authority rather than the European Commission. UK users are locked out of Siri AI on iPhone and iPad identically to EU users, despite the UK not being subject to the DMA. The available sources do not provide a conclusive explanation: Apple has not publicly distinguished its UK posture from its EU posture, and the DMCC’s enforcement timeline lags behind the DMA’s. This remains an unresolved dimension of the lockout. For the regulatory comparison framework, see the standoff article.
How did the Apple Intelligence EU delay in 2024 compare to the current Siri AI lockout?
Apple Intelligence — the broader AI feature suite including writing tools, image generation, and notification summarisation — was initially withheld from the EU at its 2024 launch under similar DMA compliance concerns. The key difference: Apple Intelligence features were discrete and negotiable, allowing Apple to reach a compliance resolution within 10 months and launch them in the EU via iOS 18.4 in April 2025. Siri AI is a deeply integrated system-level assistant whose interoperability demands are total, not feature-by-feature — which is why the 2024 negotiation pattern has not repeated. For the full precedent comparison, see the platform split breakdown.
What are the actual measurable benefits the DMA has delivered for European consumers?
Concrete DMA-driven changes include: cross-OS data transfer between iPhone and Android, browser data portability APIs enabling import and export of bookmarks and passwords, NFC access for third-party payment apps like PayPal and Vipps, browser choice screens that doubled Firefox daily active users in France and Germany, alternative iOS app stores including AltStore PAL and Setapp Mobile, and connected-device interoperability specifications for smartwatches and headphones. The Commission’s own three-year review acknowledges that impacts remain “not fully observable,” and the Free Software Foundation Europe‘s finding that none of 56 Article 6(7) interoperability requests to Apple had produced a working solution as of March 2026 tempers the compliance narrative.
How does Apple’s withholding of Siri AI compare to other DMA-driven feature restrictions by gatekeepers?
Meta and Google have both restricted features in the EU under DMA compliance pressure — Meta limited political ad targeting and Google reshuffled search result displays. The Siri AI lockout is categorically different in scale and strategic significance: those restrictions were modifications to existing features, while Siri AI is a flagship product withheld entirely from a market representing 26 percent of Apple’s revenue. The lockout also affects a product built on a competitor’s model (Google Gemini), creating a strategic irony that none of the other DMA-driven restrictions share. For the competitive comparison, see the full strategic implications.
Should app developers build for Siri AI now or wait for EU resolution?
The decision depends on user geography (what percentage of your users are EU-based?), competitive pressure (are your competitors shipping Siri AI integrations to non-EU users?), integration depth (is Siri AI core to your product or a nice-to-have?), and regulatory trajectory (is resolution likely in months, years, or never?). Teams with predominantly non-EU user bases can build for Siri AI today; teams with significant EU user bases should consider platform-agnostic AI delivery or accept an Android-first posture in Europe. The strategy article provides a full decision framework without prescribing a single answer. See the strategy and decision framework.
Where can developers submit formal interoperability requests to Apple under the DMA?
Apple maintains a formal interoperability request portal, established under DMA compliance, where third-party providers can submit requests for access to specific iOS hardware and software features under Article 6(7). As of March 2026, 56 requests had been submitted, with the Free Software Foundation Europe reporting that none had resulted in a working solution. The portal is the operational mechanism through which the DMA’s interoperability obligations are meant to function, and its track record is directly relevant to the Siri AI dispute’s trajectory. The standoff article examines the portal’s role in the broader compliance landscape.
The Siri AI lockout is a collision between two legitimate but structurally incompatible forces: a regulation designed to ensure platform contestability, and a product architecture whose privacy guarantees depend on controlled trust boundaries. Neither side is acting in bad faith, which is what makes the impasse so hard to resolve.
The practical question is what you do while the standoff persists. Start with whichever article matches what you need to understand right now. Each article stands alone. Together they give you the full picture.
