EU Orders Google to Open Android to Rival AI Assistants

The European Commission has issued two binding orders that force Google to open Android to competing AI assistants and to share its search data with rivals. Adopted on July 16, 2026 under the Digital Markets Act, they give competitors access to eleven Android capabilities that Google’s own Gemini enjoys today — from wake-word voice activation to reading what is on your screen — on deadlines that run from late 2026 to August 2028. Google says the orders endanger users’ privacy and security. Here is what was actually ordered, how Google is responding, and why none of it automatically reaches the world’s biggest Android market, India.

What Brussels actually ordered

The two measures are “specification decisions” — a DMA tool where the Commission stops waiting for a gatekeeper to comply on its own terms and prescribes, in binding technical detail, exactly how compliance must look. The first tells Google how to satisfy Article 6(7) of the DMA, the interoperability duty, for AI assistants on Android. The second specifies Article 6(11), which obliges Google to share search data with rival search engines on fair terms.

Neither decision is a fine, and no penalty was imposed for past conduct. The teeth come later: ignoring a specification decision can cost a gatekeeper up to 10% of worldwide annual turnover, rising to 20% for repeat infringements, and the Commission says it will monitor implementation for two years through regular progress reports. “Society is going through a profound digital transformation,” competition chief Teresa Ribera said in the Commission’s announcement. “We need to keep that process fair and ensure that our citizens have choice.”

Eleven doors into Android

The Commission’s case specification lists eleven features in four groups. Invocation covers launching an assistant by long-pressing the home button or navigation handle, and always-on wake-word detection, so “Hey Google” stops being the only phrase a phone listens for. Context covers centralised access to on-device app data, proactive suggestions based on what the user is doing, and “ambient data” — real-time input from the microphone, camera, screen and speakers. Actions covers structured integration with apps such as Gmail, Calendar, Maps, YouTube and Messages, background automation of on-screen tasks, and system controls like brightness and do-not-disturb. Resources covers access to on-device models such as Gemini Nano — the compact model class we examined in our Bonsai 27B explainer — plus the right to install alternative on-device models on equal terms.

The Commission’s examples are deliberately everyday: booking a taxi by voice, getting reply suggestions in a chat, asking about a place you just visited, or having a flight number surfaced mid-conversation. The point is that a rival assistant should be able to do on Android whatever Gemini can. Google’s advantage here is structural — roughly 60% of EU users are on Android devices, the Commission notes, and only Google’s assistant currently sits that deep in the operating system.

Interoperability must be free of charge, “equally effective” as what Gemini gets, and new functionality must reach third parties at the same time it reaches Google’s own assistant. For the five most sensitive features — including app-data access and screen automation — Google may run a certification program with objective, non-discriminatory criteria: draft terms are due by February 1, 2027, final terms with open applications by May 1, 2027, and Google must decide each application within four weeks.

The search data Google must hand over

The second decision reaches into the asset that made Google Google. Under the data-sharing specification, eligible rivals get the same categories of data Google uses to optimise its own results: queries, query metadata such as language and device type, viewed URLs, clicks and ranking positions. Excluded are account details, full search histories, precise timestamps, very long or rare queries, and sensitive terms — all stripped in a three-layer anonymisation process in which each user disappears into groups of at least 1,000 people sharing the same location, device and language profile.

Eligibility has a floor designed to screen out data tourists: recipients must have operated a genuine search service in the EU for two years or more (or have committed at least €50 million in capital), maintain 50,000 monthly EU users, and process the data in the EEA or an equivalent jurisdiction. Notably, AI chatbots that offer search functions qualify — not just classic search engines. The data arrives no sooner than seven days after the queries occur, access is capped at five years per beneficiary, and recipients may use it only to improve their own search technology. Training general-purpose AI models with it is expressly barred, as is systematically reconstructing Google’s results.

That is a meaningful set of guardrails, but the debate over search-query privacy is not academic — query logs are among the most revealing datasets that exist, a theme we explored in what happens to your data when you talk to an AI chatbot. Google will be audited independently before access begins, then annually.

Google’s pushback — and the Apple parallel

Google’s response was immediate and unusually sharp. “Today’s decisions risk undermining vital privacy and security guardrails for millions of Europeans,” wrote Kent Walker, Google’s president of global affairs, in a company blog post. The Android order, he argued, “threatens device security by granting external apps sensitive and powerful device permissions,” while search-data sharing means “Europeans’ private searches would be exposed to unfamiliar companies.” Android president Sameer Samat posted that the Commission “is on the wrong track”, arguing users can already switch assistants and device makers already choose what to preload.

The critics are not all on Google’s side of the aisle, and the beneficiaries are conspicuously quiet. Security outlet The Hacker News framed the order bluntly as opening the “mic, camera and screen” to third parties. The free-market ICLE argues the EU has handicapped the wrong player, since ChatGPT accounts for roughly 70% of EU chatbot use while Gemini’s Android integration was the trailing player’s main differentiator. Privacy advocates counter that gatekeepers routinely invoke security to defend distribution advantages — Gemini reached 750 million monthly active users globally in large part by riding Android. As of July 20, no rival — OpenAI, Perplexity, or any European search engine — had publicly reacted, and Google had not announced a court challenge. This is the sharpest episode yet in the fight we track in Policy & Society: who controls the AI layer on the devices people actually own.

Apple shows how this can go. Rather than grant rivals the access the DMA demands, Apple has withheld its new Siri AI from EU iPhones entirely, and on July 8 the EU General Court threw out Apple’s challenges to its gatekeeper designation — a ruling that also closed the door on pre-emptive legal blocks against exactly the kind of specification decision Google just received. Google could follow Apple’s playbook and degrade or delay Gemini-class features in Europe; an appeal, if it comes, would not suspend the obligation to comply.

The India angle

Nowhere outside Europe watches Google-Android regulation more closely than India, where Android runs 93.05% of mobile devices as of June 2026, per StatCounter. Yet these decisions stop at Europe’s border: the Commission’s announcement states the measures apply to European users and the EU market. Google’s record says it takes such borders literally — it recently removed the search choice screen in Switzerland while keeping it in the EU, prompting a preliminary probe by the Swiss competition authority on July 14.

India has its own, older Google-Android fight. The Competition Commission of India fined Google ₹1,337.76 crore in October 2022 for abusing Android dominance through mandatory app bundling and restrictive agreements; the NCLAT upheld the penalty in March 2023, and Google’s further appeal went to the Supreme Court, where no final verdict has been reported as of publication. The remedies Google did implement in January 2023 — an India-specific choice screen, unbundled app licensing, easier forks — came from Indian orders, not European ones. Meanwhile India’s DMA equivalent, the Digital Competition Bill, has been paused since a parliamentary committee recommended reconsidering it in August 2025, with the corporate affairs ministry now re-studying its thresholds. Our read: there is currently no live Indian legal mechanism that would force Google to replicate the EU’s AI-assistant access or search-data sharing here — Indian users get these changes only if the CCI, the courts, or a revived bill acts.

The demand side, though, is very much present. A YouGov survey this month found 57% of Indian respondents already use AI assistants or chatbots for online discovery, against 65% for search engines — a narrower gap than in most markets. If assistant choice becomes a real differentiator on European Android phones, Indian consumers will notice what they are not getting. European court validation of Android-dominance findings has also strengthened Indian regulators’ hand before, as coverage of this month’s EU fine ruling noted.

What to watch

The compliance calendar is the story now. By end-August 2026 Google must open eligibility applications for search data; template licences and test datasets follow in September, the finalised dataset in November, and a pricing offer by January 2027. On the Android side, certification terms land in February and May 2027, the main features ship with Android 18 by August 1, 2027, and simultaneous wake-word listening for multiple assistants arrives with Android 19 by August 2028. Watch four things: whether Google files a General Court appeal it has so far only hinted at; whether it ships these capabilities EU-only, as the Switzerland episode suggests; whether OpenAI, Perplexity or a European search player actually applies — silence so far is the most interesting data point; and whether a separate, reportedly imminent DMA fine over self-preferencing lands before the Commission’s summer recess. Any of the four could reshape how much of this ambitious order becomes lived reality on Europe’s phones — and how loudly India’s regulators take notes.