EU Orders Google to Open Android to Rival AI Assistants

By Suad Seferi · Aug 12, 2026

Google will have to give rival AI assistants deeper access to Android, breaking down advantages that currently allow Gemini to use parts of the operating system that competitors cannot reach. The European Commission adopted binding measures on 16 July under the Digital Markets Act requiring Google to open 11 Android features used by AI assistants to competing services. The rules are intended to let alternatives to Gemini interact with Android more like Google's own assistant does. For users, that could eventually mean choosing another AI assistant and still being able to activate it by voice, let it understand what is happening on the screen, interact with installed apps and carry out tasks across the phone. Google has until the next major Android release, Android 18, and no later than 1 August 2027, to implement most of the measures. One of the most technically sensitive requirements, allowing several AI assistants to listen concurrently for their own wake words, has a later deadline. Google must introduce that capability with Android 19 by 1 August 2028. The decision turns the growing competition between Gemini, ChatGPT, Claude and other AI services into an operating-system battle. Until now, installing another AI chatbot on Android has not necessarily given it the same access that Gemini receives. Brussels wants that to change. More than replacing one chatbot with another The Commission's decision goes much further than allowing users to select a different default assistant. Rival AI services will be able to request access to capabilities covering four broad areas: activating the assistant, understanding a user's context, taking actions across applications and accessing Android's computing resources. One example is voice activation. Today, Google's own services can respond to “Hey Google” even when a device is locked, idle or running in power-saving mode. Under the new measures, users will eventually be able to activate competing assistants with their own wake words under similar conditions. The rules also cover one of the most valuable features in the emerging generation of AI agents: the ability to act inside other apps. A competing assistant could, with permission, draft and send an email using the user's chosen email application, create a note, schedule a meeting or add items to a shopping list. It could also automate longer tasks. The Commission gives the example of asking an assistant to order groceries. The AI could read a shopping list, open the relevant supermarket app and add the items in the background while the user continues doing something else. Those capabilities move AI assistants closer to becoming an interface for the entire phone rather than another app sitting on the home screen. ChatGPT or Claude could get much deeper Android access The Commission does not dictate which companies must use the new access. But the framework creates an opening for developers of competing assistants, including services from OpenAI, Anthropic and other AI providers, to build much deeper Android integrations if they choose to do so. The significance is that Google will not be allowed to reserve covered operating-system capabilities for Gemini. Among the 11 features are access to screen context, selected sensor information, background execution and structured integrations that let assistants perform actions inside applications. Rival services must also be given equivalent access to certain on-device AI resources available through Android. That includes access, under equivalent conditions, to on-device models such as Gemini Nano when those models form part of the Android operating system covered by the Digital Markets Act decision. Third-party developers will also be able to install and run their own on-device models under hardware and background-processing conditions comparable to those available to Google's models. For smaller AI companies, that could matter as much as access to the screen or microphone. Building an assistant that feels native to a smartphone requires more than a capable language model. It needs operating-system access, reliable background processing and a way to interact with the user's applications. Those are precisely the layers the Commission is opening. Google says Europe is creating a security risk Google strongly opposes the decision. Kent Walker, Google's president of global affairs, said the measures could weaken privacy and security protections by granting external applications access to sensitive Android capabilities. Google argues that third-party AI assistants already have routes to Android features and that phone manufacturers play an important role in determining which services should receive deeper access. The company also says Brussels has underestimated the risks created when external AI applications receive powerful permissions. The Commission has built safeguards into its decision. Users must consent before an AI service receives access to protected features. For particularly sensitive capabilities, Google can also impose objective security, privacy and integrity requirements on third-party providers, provided those conditions are non-discriminatory and not used as a commercial barrier. Independent third parties will be involved in certifying whether services meet those requirements. Google must publish draft terms for that eligibility programme by 1 February 2027, finalise them by 1 May 2027, and begin accepting certification applications from that date. The disagreement is therefore not over whether security controls should exist. It is over how much control Google should retain over those controls when Gemini is competing with the same companies seeking access. Android is becoming the next AI distribution battle The stakes are considerable. The Commission estimates that roughly 60% of European mobile users are on Android devices, giving Google a powerful distribution channel for Gemini. It argues that this advantage becomes difficult for competito…