Google challenges EU order to open Android to AI rivals

Google has filed a legal challenge against two European Commission orders that would force it to open Android and Search to AI competitors, taking the case to the EU General Court in Luxembourg. The move targets July 2026 decisions requiring Google to give rival AI assistants access to Android features comparable to what its own services receive, plus a separate mandate to share search data with competing search engines, according to The Irish Times and The News International.

This isn't a fight over whether apps like ChatGPT, Claude, or Perplexity can run on Android at all. They already do, installed as ordinary sandboxed apps that Google says are "used by tens of millions of consumers daily," a figure that comes from Google's own security blog rather than independent usage data. What's contested is whether qualifying rival assistants can get access to invocation, hardware, and system controls comparable to what Google's own services receive, not whether they can exist on the platform in the first place.

Google challenges EU order as Android AI access deadlines approach

On July 16, 2026, the Commission adopted binding decisions ordering Google to provide competing AI services with "free and effective interoperability" across 11 specific Android features, on terms equal to what Google's own apps receive under Article 6(7) of the Digital Markets Act, according to ECIPE. Part of that order requires Google to let users select and activate a preferred third-party AI assistant through voice commands within 12 months of the decision, The Irish Times reported, a timeline that lands close to, but isn't stated as identical to, the broader Android 18 deadline described below.

ICLE's analysis of the Commission's earlier April 2026 draft specification, which hasn't been matched line-for-line against the final July decision, describes a fuller list of capabilities rival assistants are meant to gain: background access to the microphone, camera, and GPS, custom wake-word detection, screen automation, OS-level controls, and access to on-device AI hardware including Gemini Nano's NPU, per ICLE. That caveat matters here: draft language and finalized requirements don't always match, so readers weighing how far this actually goes should check the Commission's published decision documents as they surface.

It helps to separate what's already available from what's still being litigated. Installing a rival AI app on Android is already possible and always has been. Selecting that app for voice activation, so saying its wake phrase launches it, is the near-term requirement due within 12 months of the July decision. Granting it system controls, like background sensor access or screen automation, and letting it run alongside "Hey Google" through concurrent wake-word detection, are separate, later-stage requirements with their own deadlines. None of those last three exist on any phone yet.

The compliance clock reflects that split. Most changes are due in Android 18 by August 1, 2027, while concurrent hotword detection, letting a rival wake word work alongside "Hey Google" rather than replacing it, must be implemented in Android 19 by August 1, 2028, per ECIPE. The Commission's reasoning leans on scale: roughly 60% of European mobile users own an Android device, and its preliminary findings describe that footprint as "a important gateway" for standalone AI services trying to reach end users, per ICLE.

Google says the rules weaken Android security; the Commission says safeguards are built in

Google's appeal argues the orders would force it to hand over access without adequate safeguards and would undermine Android's security architecture. "We're appealing decisions that will force us to share people's private search history without sufficient anonymisation and weaken vital security protections on Android," said Oliver Bethell, Google's senior director for competition, according to The News International.

In an August security post, Google went further, claiming the Commission's proposed Trusted Certification Authorities program would let apps gain access to sensitive features through third-party certifiers, with no direct oversight from Google or device makers. The post also points to indirect prompt injection as a documented risk in agentic AI systems, per Google Security Blog. That post was co-signed by researchers Google selected, not an independent audit, so it reads as Google's own characterization of the risk rather than settled proof of it.

According to ECIPE's account of the specification, the Commission disputes that framing. Access to each feature requires explicit user consent, and for five particularly sensitive functions, including screen automation and centralized on-device app-data access, Google may impose objective, non-discriminatory eligibility conditions backed by independent certification, per ECIPE. The Commission has separately said both measures "contain strong safeguards protecting the privacy of users, device integrity and security," according to The News International.

DMA Article 6(7) already permits integrity measures that are "strictly necessary and proportionate," per ICLE. Whether Google's specific objections meet that standard is now a matter for the General Court to sort out, not something either company's or regulator's public statements settle on their own.

The Google search data sharing order

The Commission's framework describes the data covered here as anonymized. It is not described in the sources reviewed as a handover of identifiable search histories.

Under Article 6(11), Google must give qualifying search competitors, including AI chatbots with search features, access to anonymized ranking, query, click, and view data. That access has to run on fair, reasonable, and non-discriminatory terms, per ECIPE. By January 2027, rival search engines are meant to gain access to scale-level data, particularly from AI chatbot interactions, that currently only Google collects at that volume, regulators said, according to The Irish Times.

Access isn't open to anyone who asks. Recipients generally need at least 50,000 monthly EU users, or, for companies founded within the past two years, more than €50 million in funding, per ECIPE. The data also comes with restrictions: it can't be used to train general-purpose AI models, build consumer profiles, sell advertising, or feed products unrelated to search, which the Commission frames as its core safeguard against misuse.

The Commission's timetable for this order began phasing in from August 2026 and runs through January 2027, but the supplied reporting doesn't establish which recipients, if any, have actually received data yet or whether every milestone has been completed. Google's appeal challenges this decision on the same privacy grounds it raises against the Android order. Bethell put the stakes in personal terms: "People use Search for their most personal questions, from medical worries to close relationships, and mandating we share these personal queries without adequate safeguards would cause irreversible harm to user privacy," per The Irish Times.

What Android and search users can expect right now

No confirmation exists in the current reporting that any rival assistant has gained voice-activation status on Android, or that any search competitor has started receiving data under the new framework. Nothing in the sources reviewed identifies a new setting, toggle, or rollout on a typical Android phone, or in Search, today.

The August 2027 and August 2028 Android deadlines, along with the search-data timetable running through January 2027, are the Commission's announced implementation schedule. What the research doesn't establish is whether the General Court's review pauses that schedule, whether either side has sought interim relief, or whether Google has met any milestone so far. That's a question for the court, not for either party's public statements.

The research also doesn't specify which Android device models or manufacturers would carry the responsibility for supporting these features, how a rollout would actually work across different EU markets, or which specific companies have applied for search-data access beyond meeting the stated eligibility thresholds. Those gaps matter because the dates in circulation describe a legal deadline, not a feature that's been built, tested, or shipped on any device yet.

Google isn't the only company navigating this rulebook. Apple has said Siri's AI features won't launch with iOS 27 or iPadOS 27 in the EU, citing DMA interoperability requirements, according to ECIPE's account of Apple's announcement, a separate decision from Google's case rather than a direct consequence of it.

No user-facing change is identified in the supplied reporting. The developments worth watching next are the General Court's proceedings in Luxembourg and whatever implementation updates the Commission publishes as the 2027 and 2028 deadlines get closer, not a new toggle to look for on a phone today.