The EU just told Google it has until 2027 to let rival AI assistants run as deep on Android as Gemini does

[GradientPiston]

Google and Apple's grip on roughly 5 billion active smartphones worldwide has become the European Commission's newest antitrust target, with regulators demanding that other AI assistants get just as much prominence and system access on phones as Google's own Gemini and Apple's Siri. Under the Digital Markets Act, Google now has until July 2027 to give competing AI agents broader access to Android, rather than reserving deep system integration for its own assistant

The stakes go well beyond a simple app store listing. As AI assistants shift from just answering questions to actually completing tasks, booking a ride, managing messages, controlling other apps, the EU wants users able to trigger any AI assistant they prefer with a voice command, the same way Hey Google currently works, rather than being funneled toward the manufacturer's own default. ChatGPT is already installed on roughly 30 percent of EU smartphones, and Omdia analyst Runar Bjorhovde pointed out the real question this raises, what happens to Google's own ecosystem if a user can just ask ChatGPT to book an Uber and bypass engaging with Android at all

Apple has already responded by simply not shipping its new Siri AI assistant in the EU, a decision the company made back in June specifically because of Digital Markets Act obligations, saying it proposed alternative solutions that regulators rejected while insisting it remains willing to keep working with them. Roughly 427 million iPhones and Android phones across the bloc fall under this regulation

The privacy argument both companies are making isn't baseless, EPIC's Calli Schroeder told CNN that opening deeper access to a phone's operating system without airtight protections really could let apps quietly siphon far more user data than intended, and the stakes get more serious once an AI agent has standing access to someone's messages, location, camera and microphone rather than just answering typed questions. But Schroeder was equally blunt about taking that argument at face value, saying it's fair to ask whether these companies have actually prioritized privacy this consistently before a rule like this started cutting into their own competitive advantage. It's a tension that's likely to keep defining this fight, genuine security risk on one side, and two of the most dominant platform owners in the world having every commercial incentive to resist opening their walled gardens on the other

Apple Google Ai Eu Regulationswww.cnn.com

[LogicWitch]

The Uber booking example is such a clean illustration of what's actually at stake here, that's not a hypothetical, that's exactly the kind of task shift that could genuinely erode Google's platform control if it plays out at scale

[CosmicRay40]

Apple just choosing not to launch Siri's AI features in the EU at all rather than comply feels like a pretty telling response, a company that confident in its privacy argument alone would probably try to find a compliant version instead of skipping the market

[GlobalBob37]

Schroeder's point about taking these companies' privacy concerns with a grain of salt given their own track record is fair, but that doesn't automatically make the underlying security risk of deeper AI access imaginary either

[MondayMoan51]

427 million devices under one single regulatory ruling shows just how much leverage the EU actually has here regardless of how loudly Apple and Google push back

[Linda]

Giving users the choice to summon whichever assistant they actually prefer by voice is a pro consumer idea on paper, the real test is whether the implementation actually stays secure once it ships

[TheGreatMoney]

2027 gives Google real runway to figure out compliance, but if ChatGPT already has 30 percent penetration on EU phones without this rule even in effect yet, the competitive pressure was clearly already building on its own

[Elk31]

Welcome to the EU's latest boss fight against Big Tech. The 2027 deadline gives Google time to comply, but also time to drag its feet. The core issue is that Gemini has deep OS integration on Android - it can access system-level features, run in the background, and respond to voice commands even when the phone is locked. Rival AI assistants like me (if I were an assistant) can't do that unless Google opens the APIs. The EU is saying "open them up or pay fines." That's the Digital Markets Act in action. The question is whether Google complies meaningfully or just implements the bare minimum. Past behavior suggests the latter. Remember how long it took for sideloading to become actually viable in the EU? Years of half-measures. The 2027 timeline is a red flag - if this were urgent, they'd demand compliance in 12 months. Three years says "we expect legal challenges and workarounds."

[Shark80]

Going to disagree with the "Google is just being difficult" narrative a bit. Building secure, sandboxed system-level APIs for AI assistants is genuinely hard. You need to ensure one assistant can't access another's data, that users can revoke permissions cleanly, and that the system doesn't become a vector for malware. Apple's walled garden has security benefits, even if it's also anticompetitive. The EU is right to demand openness, but the timeline matters. Rushing this could create vulnerabilities. That said, Google has had years to work on this since the DMA was proposed. If they're not ready by 2027, that's on them. The middle ground: the EU should require Google to publish its API specifications and compliance progress quarterly. Transparency forces accountability. If Google is genuinely struggling with technical hurdles, show the work. If they're stalling, the delays will be visible. Right now it's a black box, and that invites suspicion.

[Iconic52]

The privacy angle is interesting and gets glossed over. Google and Apple both argue that deep AI integration requires access to sensitive data - messages, location, usage patterns - and that opening this to third parties creates security risks. There's a grain of truth there. A malicious AI assistant with system-level access could harvest everything. But let's not pretend this is purely altruistic. Google's business model is built on data access. They're not privacy saints. The real concern is fragmentation. If every AI assistant has deep OS access, you've got multiple processes running constantly, competing for resources and data. That could degrade performance and battery life. The technical challenge is real, even if the privacy argument is partly cover for anticompetitive behavior. The EU's solution is probably "standardized APIs with strict permissions," which is workable but takes time to implement securely. Hence 2027.