EU Forces Google to Share Search Data and Open Up AI on Android — What It Means for AI Competition

Android powers roughly 60% of European mobile users. Starting August 2027, every one of those devices must let users pick any AI assistant as their default, with full access to the microphone, camera, on-screen content, and the ability to execute background actions just like Google's own Gemini. The European Commission adopted two binding specification decisions under the Digital Markets Act on 16 July 2026 that force Google to open its mobile platform to rival AI services in ways that have no precedent in the history of mobile computing. These are not fines. They are structural mandates that rewrite how AI assistants compete on the world's largest mobile operating system.

The first decision forces Google to give third-party AI assistants the same system-level hooks that Gemini already enjoys. The second requires Google to hand anonymised search query, click, and ranking data to competing search engines and AI chatbots at a cost-based fee. Taken together, the rulings represent the most aggressive regulatory intervention in AI platform competition since the DMA itself was conceived.

What the Android Decision Actually Requires

The Commission's Android specification covers 11 operating system features. Five are designated as restricted features that require certification through a new Qualified AI Assistant Programme. These include centralised on-device data sharing (AppSearch), context-aware intelligence (the always-on machinery behind features like Magic Cue), structured on-device integration (App Actions and App Functions), screen automation (Computer Control), and deep system integration covering settings, media, screenshots, notifications, and power management. Google can require assistants to pass certification by a Trusted Certification Authority before touching these five features, but it must let third parties set up those certifiers on reasonable and non-discriminatory terms.

The other six features carry no certification requirement at all. These include ambient data access (microphone input, system audio, camera, screen contents, location, and sensor data), always-on hotword detection on the low-power DSP, long-press invocation, system-level on-device AI models, third-party model implementation, and background execution. Paragraph 119 of the decision opens these to all third-party apps and bars Google from restricting the type or use case of the app that calls them. The practical effect is seismic: any AI assistant a user installs can listen for its wake word with the screen off, see whatever is on the display, read and draft emails in Gmail, create Calendar events, pull content from Drive and Docs, trigger Maps navigation, control YouTube playback, read and write SMS and RCS messages, and place phone calls.

Google's first-party services currently reach these sensors with reduced consent processes and lighter privacy indicators. Third parties get runtime consent per use. The decision equalises both sides: third-party assistants must get the same lightweight consent prompts that Google's own services use, and Google cannot impose extra friction on rivals that it does not impose on itself.

The Search Data Mandate and What It Opens

The second decision targets Google's search monopoly as it intersects with AI. Google must now hand anonymised search query, click, and ranking data to competing search engines and AI chatbots at a cost-based fee. For founders building AI search products or retrieval-augmented generation systems, this is the regulatory equivalent of a gold rush. Google's search signals have been the single most valuable training and ranking dataset in the industry, effectively locked behind Google's own products. The DMA decision breaks that lock, giving startups access to query patterns, click-through rates, and ranking signals that were previously impossible to obtain at any price.

The cost-based fee provision prevents Google from pricing the data prohibitively. The Commission will oversee the fee structure to ensure it reflects the actual cost of providing the data, not the value of the data itself. That distinction matters: Google cannot charge what the data is worth to competitors. It can only charge what it costs to package and deliver it.

Timeline and Enforcement Structure

Google must ship the Android changes in the next major release, Android 18, and by 1 August 2027 at the latest. The multi-assistant listening feature (multiple wake words from different assistants on the same device) slips to Android 19 with a deadline of 1 August 2028. The Qualified AI Assistant Programme must be operational before Android 18 ships. Trusted Certification Authorities will certify third-party assistants free of charge, and Google cannot revoke an assistant's certification once issued. It can revoke a TCA's certification authority, but only after clearing the change with the Commission two months in advance. The Commission's separate power to open non-compliance cases, including fines, remains untouched and independent of these specification decisions.

For the search data mandate, the timeline is shorter. Google must begin providing the data within six months, with the cost-based fee structure subject to Commission approval. The data covers anonymised query logs, click-through data, and ranking signals across all search verticals where Google holds gatekeeper status.

What This Means for Founders

Three strategic implications stand out. First, Android becomes the largest distribution channel for AI assistants that is not controlled by a single AI company. If you are building a vertical AI agent for healthcare, finance, education, or any domain where voice or screen-aware interaction matters, getting default placement on Android's AI assistant choice screen could be your single largest customer acquisition channel. The browser choice screen precedent from the earlier DMA enforcement gives a sense of the potential: when Android users were asked to pick a search engine, Google's market share dropped and competitors saw meaningful gain. The AI assistant choice screen could have a similar effect at a much larger scale.

Second, the search data sharing provision changes the competitive landscape for AI search and RAG products. Startups that could not afford to build search infrastructure from scratch can now access Google-scale signals at cost. The caveat is that the data is anonymised and aggregated, which limits some use cases, but even anonymised query and click data at Google's volume represents a step-change in what independent search products can achieve.

Third, the consent equalisation creates compliance requirements but also competitive opportunity. Any AI assistant that meets the certification bar can reach into the same system features that Gemini uses. The certification programme itself creates a new market: Trusted Certification Authorities that validate AI assistant security and privacy practices will be in demand, and the bar for certification will effectively define the baseline security standard for mobile AI assistants across Europe.