Under Article 6(7) of the DMA, Google must provide developers with free and effective interoperability with hardware and software features controlled by its operating system, Google Android.
To assist Google in enabling such interoperability, the Commission opened specification proceedings on 27 January 2026. These proceedings aim to specify measures that Google must implement to ensure effective interoperability under the DMA with 11 Google Android features relevant for AI services.
The final decision was adopted on 16 July 2026 and includes the measures that Google must implement to enable effective interoperability.
Background on the proceedings
Why did the Commission decide to specify how Google must provide interoperability with Google Android features?
The DMA requires Google to grant competing AI services the same access to Google Android operating system (OS) features that Google’s own services have. The Commission’s decision seeks to ensure that Google enables effective interoperability for alternative AI providers, in compliance with the DMA.
AI assistants are becoming a primary way for users to interact with their mobile devices. Gemini, backed by Google’s full‑stack AI capabilities, is uniquely placed to become the leading AI offering on mobile devices.
With around 60 % of European mobile users on Google Android devices, such devices are a key channel for AI services to reach users. Google already uses this extensive distribution network to deliver its AI services, including Gemini, to users. Google also ensures that Gemini gets deep level access to key operating system features, allowing it to offer users truly attractive assistive experiences. At the same time, many of these key features, such as invoking the assistant through voice command, or taking actions on other apps and understanding users’ context, are almost exclusively available to Gemini, with user-installed AI assistants prevented from accessing these functionalities.
Genuine competition and innovation can only develop if alternative providers of AI services can offer users a similarly powerful assistant as Google does. This requires a deep level of access to the Google Android operating system on par with that of Google’s own AI services.
Benefits for businesses and users
How will these measures benefit businesses and users?
As a result of specification proceedings, third-party AI assistants and AI-powered services will be able to effectively access 11 Google Android features that are crucial for providing a feature-rich and performant service.
These features will enable third party providers of AI services to deliver a wide range of user experiences, without sacrificing safety or performance. These user experiences may include:
- The ability to easily activate by voice the user’s preferred AI service for a specific task very much like “hey, Google”;
- The ability to instruct the AI service to perform actions across the user’s apps on the device, such as drafting and sending an email via the user’s preferred email app or adding items to a shopping list;
- The ability to benefit from AI-powered live translation tailored to a variety of languages, or from speech-to-text recognition;
- Receiving “proactive” suggestions for relevant actions or information in apps, without the user having to actively prompt the AI service, similar to what Google already offers. Such experiences include the user being shown the flight number stored in its email app, when their partner asks about it via a message or when the flight is nearing;
- The ability to automate tasks, such as ordering food from a restaurant, where the AI assistant handles all the necessary steps – including opening the food delivery app, interacting with it to select the right dish (possibly with a prompt “as usual”), and adding the delivery details.
On Android, these experiences are currently reserved to Gemini as other assistants do not have access to the operating system features which would allow to perform these tasks.
Developers whose apps may be used by an AI service, such as email apps, ride sharing apps, or messaging apps for which the user asks the AI service to perform a specific task will also benefit from more engagement and discoverability through integration with alternative AI services.
Which features of Google Android will be made interoperable thanks to these measures?
The Decision covers 11 features of Google Android, which represent the high-level building blocks that AI services rely on:
- Invocation, or the ability for a user to start their interaction with an AI service through a variety of access points, such as the home button;
- Context, or the ability for an AI service to gather data (including from apps, sensors or the screen) to then understand and anticipate the user’s needs;
- Actions on apps and the OS, or the ability for an AI service to execute tasks on a user’s behalf; and
- Access to resources, or the ability for an AI service to use sufficient hardware and software resources (including calling Google’s on-device AI models which are part of the DMA designated operating system) to execute its tasks.
The user will always be able to explicitly consent to access for any of the features, for the AI assistants they choose to install.
These features are explained in more detail below:
Features for invocation
- Long-press home button/navigation handle
- The user will be able to invoke a third-party service by pressing the centrally located home button (or handlebar) to offer services based on contextual data.
- Access points can no longer be reserved for Google services, such as Google Search (via Circle to Search).
- Always-on hotword detection
- A user will be able to activate a third-party service by saying a “hotword” or wake word. Users will be able to activate their preferred AI assistant at any time, such as when driving, cooking, or listening to a podcast on their headphones, and even if the device’s display is turned off, or the device is in battery saving mode or in standby.
- The measures will also enable third parties to access this feature on a concurrent basis, allowing users to use different AI services for different tasks.
- Hotwords can no longer be reserved for Google services, such as Google Assistant or Gemini (via “Hey/OK Google”).
Features for context
- Centralised access to apps’ data stored on-device
- Alternative AI services will be able to access data that applications decided to store on device and that those apps and the user have chosen to share with other AI assistants. This feature enables access to take place in a centralised manner instead of on an app-by-app basis, in the same fashion as Google services currently do via AppSearch.
- Storing data on-device has several advantages compared to on-cloud storage, including enhanced privacy, shorter response time, more efficient cross-app search and retrieval capabilities, and the ability to access the data without the need for an internet connection.
- Centralised access enables, for instance, AI services to autonomously plan and perform actions, benefitting use cases that require quick access to the user’s context.
- Context-aware intelligence
- Third parties will be able to offer AI services that are personalised to the user's context, continuously present to enhance the device experience and capable of suggesting actions based on the user’s needs, without the user having to actively invoke the assistant (i.e. proactive actions), with the user’s consent.
- This feature will, subject to user consent, enable third parties to access multiple functionalities, including input data such as physical and digital context (e.g. screen contents or device sensor data such as location signals) but also processing resources to turn these inputs into proactive suggestions or intelligent experiences, and outputs or “surfaces” to display these experiences.
- For instance, alternative AI providers will be able to offer live translation experiences, show relevant information to users such as their flight booking number during a phone call or proactively suggest a restaurant when chatting with a friend.
- This will enable third-party AI services to benefit in the same way from these capabilities, which are already used by Google’s services through, for instance Magic Cue’s proactive suggestions.
- Ambient data
- Ambient data is the continuous stream of real-time inputs/outputs from a device's core sensors (e.g., microphone, camera, screen, speakers). It enables AI services to deliver personalised, context-aware experiences such as sound detection, live screen guidance, or object recognition by processing raw sensor data seamlessly.
- Third-party AI services will have the same access as Google does to real-time sensor inputs including microphone, camera, screen, and speakers, under the same conditions for user consent and awareness as apply to Google.
Features for actions on apps and the operating system (OS)
- Structured on-device integration
- AI services will be able to easily interact with other apps installed on the device and perform tasks on behalf of the user within those apps, for tasks that the apps and the user have chosen to make available to AI services. These tasks include “send a message”, “create a note”, “schedule a meeting”. This includes access to certain Google apps (i.e.Gmail, Calendar, Drive, Docs, Maps, YouTube, Messages and Phone) that Alphabet will make available through operating system-level integration channels.
- This capability is essential for AI services, as it allows them to become a truly capable assistant for the user across the ecosystem of applications on its device.
- For instance, Android implements structured on-device integration through App Functions, which developers can enable for their apps, and which can be accessed by AI services without being reserved anymore for Google services, such as Google Assistant or Gemini.
- Screen automation
- AI services will be able to automate multi-step tasks within apps, on behalf of the user upon their consent. They will do so by imitating user behaviour in a separate virtual window, which makes it possible for the assistant to complete the task in the background, while the user can do something else. For example, the user can task an AI service with ordering the items from their shopping list from the local supermarket. Screen automation allows the controlling app to open the shopping list app in the background, look at its contents, and then open the app of the supermarket to order each item from the list.
- Android implements screen automation via Computer Control, which can automatically access apps, and which is currently reserved for Google’s services, such as Gemini.
- System integration
- It will enable for third party AI services seamless integration with the operating system and its settings, such as changing the display brightness, interacting with playing media, enabling do not disturb mode, or turning off Bluetooth.
Features for access to resources
- System-level on-device models
- AI services will be able to call on existing on-device models (“ODMs”), including the Gemini Nano ODMs, that are part of the DMA designated operating system, already preinstalled on Android devices and already made accessible to third parties. As a result of the measures, third-party AI services will have guarantees of equal access (for example, in terms of performance) to ODMs, as Google’s services.
- On-device models are machine learning models that run directly on the device hardware. As a result, ODMs can process data locally instead of via cloud, providing shorter response times, offline availability, and enhanced privacy protection. Access to ODMs can be used by users for summarizing text, proofreading and speech recognition.
- On-device model implementation
- Third parties will be able to install, run and use on-device models (ODMs) under the same hardware‑resource and background‑execution conditions that Google’s own models enjoy, and will allow their ODMs to be shared centrally with other apps.
- Choice and innovation regarding ODMs is important, because these models have different strengths, such as size, speed, or accuracy. For example, ODMs can go beyond just text and can be capable of ingesting and outputting images, sound, or video.
- Background execution
- Background execution allows an app to execute actions in a timely mannerin the background, including when the user is not actively looking at the app or is not using the device. This is particularly useful for AI-related apps, which in many cases are designed to help the user while performing other tasks in other apps, or even to automate tasks while the phone’s screen is off.
- Access to background execution will bring benefits to developers, enabling third-party apps to have the same access to background execution as Google’s apps through the application of transparent, objective, precise and non-discriminatory rules.
General measures for all features
- Google will have to enable interoperability, free of charge, for all the features, throughout the Google Android ecosystem, including on Android devices from other manufacturers.
- The interoperability solutions will have to be equally effective to those available to Google and must not imply unnecessarily cumbersome settings, additional user friction or be conditioned upon the AI provider holding a default role.
- Google will have to provide third parties with complete documentation of the interoperability solutions, allow them to test the interoperability solutions, and provide the necessary technical assistance.
- Google will have to make available to third parties any new functionalities of all features, once they become available to Google.
Timeline
By when should Google implement the measures for each feature?
Google will have to implement these measures in the next major Android release, which is Android 18, and by 1 August 2027 at the latest. Concurrent hotword detection allowing multiple services to be triggered via voice invocation must be implemented in Android 19, that is by 1 August 2028 at the latest.
Implementation of the measures
Do the measures contain any safeguards to protect privacy, security and integrity?
Under the DMA it is gatekeeper’s responsibility to implement strictly necessary and proportionate measures to safeguard the integrity of their systems. In addition, the whole body of EU regulations setting the standards of privacy and security, in particular the GDPR and Cyber Resilience Act continue to apply.
In addition, through user consent mechanisms, users are to decide which AI services to use – including those that offer improved privacy and security – and which AI services can access particular features on their devices.
In exceptional circumstances, in view of the sensitive nature of some of the features, Google may put in place objective and non-discriminatory eligibility conditions to limit access to third parties meeting certain privacy, security and integrity standards. No further commercial requirements may be imposed. These features include:
- screen automation;
- structured on-device integration;
- system integration;
- centralized access to apps’ data stored on the device;
- context-aware intelligence.
Along with Google, independent third parties will certify that a third-party app meets such criteria. By 1 February 2027, Google must publish the draft terms for the eligibility program for consultation by third parties and by the Commission. By 1 May 2027, Google must publish the final terms of the program. As of 1 May 2027, Google must accept applications for certification. The assessment of whether a specific AI service meets the criteria must be finalized within four weeks upon receipt of the application. Non-AI services will be able to request access to the features via a separate process.
The Android ecosystem relies on many device manufacturers that implement Android on their mobile devices. Will the measures prevent them from innovating and differentiating their devices?
The measures do not undermine device manufacturers’ ability to customise and differentiate their products. Manufacturers remain free to customize Google Android and the features as long as these modifications do not impede effective interoperability.
In particular, manufacturers remain free to differentiate their smart mobile devices through various means, including pre-installing certain apps, setting default apps in an out-of-the-box experience (and thus monetise that promotional space), offering innovative complementary software and user interfaces, leveraging hardware features such as cameras and batteries, and maintaining their branding and product identity.
Will the draft measures increase device costs to the detriment of EU consumers, due for example to more expensive chipsets for devices?
The proposed measures do not mandate anything beyond what the devices can already technically sustain. For example, the measures concerning the enabling of voice activation of an AI assistant will only concern devices already equipped with the necessary chip. Device manufacturers will therefore not be required to equip their devices with any additional hardware susceptible of driving prices up.
Under the proposed measures, the implementation of the interoperability solutions is on Google. Google cannot shift the implementation effort on device manufacturers.
Given how quickly AI assistant features develop on mobile devices, will the measures prevent Google from changing or adding new features so as to escape compliance?
The decision requires Google to ensure effective interoperability with the 11 features described above. The decision does not specify the exact technical implementation.
If Google adds new functionalities to any feature covered by the decision, these must be made available to third parties at the same time when they become accessible to Google’s own (AI) services on the device. For any new features not captured under the decision, Google remains responsible to ensure full and effective compliance with its interoperability obligations under the DMA.
Will interested developers be able to test the features before offering them to users?
Yes. Google must provide reasonable technical assistance and allow developers to test the interoperability solutions (including via beta testing).
Has the Commission engaged with Google and third parties?
Yes. Google and the Commission are in regular dialogue on Google’s compliance with the DMA. Google and the Commission have extensively engaged during these specification proceedings, including through numerous technical meetings and discussions, including between Google’s and the Commission’s technical experts and engineers.
The Commission also regularly met with interested third parties and gathered feedback via a public consultation. These exchanges fed into the final measures, which take due account of Google’s and third parties’ observations.
Will the Commission monitor how Google implements the Decision?
The Commission will monitor Google’s implementation of the Decision, in particular over the coming two years, during which Google must regularly report to the Commission on the progress of designing, developing, implementing and releasing the interoperability solutions.
Will the measures and the specification decision be published?
Yes, both the non-confidential version of the measures and of the specification decision will be made publicly available on the Commission’s website (DMA.100220). This ensures transparency regarding the measures Google must implement.
The measures list the concrete actions that Google must take, while the decision describes the Commission’s reasoning for adopting these measures, how the features work, and why Google must open them up. The measures should be read together with the decision to understand how the Commission decided on the measures.