Disclaimer
This article is based on the European Commission’s official announcement and publicly available reports. It was written as of July 19, 2026, and includes analysis and interpretation by DANA NOTES.
What This Article Covers
News Summary
On July 16, 2026, the European Commission adopted two binding specification decisions concerning Google under the Digital Markets Act, or DMA.
The first decision requires Google to open key Android operating system features so that rival artificial intelligence services can operate under conditions more comparable to those available to Gemini.
The second decision requires Google to provide rival search engines with anonymized search data collected at scale by Google Search. AI chatbots that offer search functionality may also be eligible to receive the data if they meet certain requirements.
Rival AI assistants can currently be installed on Android devices, but they have limited access to core operating system functions such as voice activation, screen context, app control, and background tasks.
The EU concluded that these differences make it difficult for AI services to compete fairly.
Two Measures the EU Requires Google to Implement
The decisions are divided into two main areas: ensuring interoperability with Android AI features and sharing Google Search data.
1. Opening Android AI Features
Google must provide technical access that allows rival AI services to use major Android features at a level similar to Gemini.
The European Commission identified 11 Android functions that are necessary for AI services.
The key functions include:
- Activating an AI assistant through a button or voice command
- Understanding user context through the screen, sensors, and app data
- Performing actions in other apps or through the operating system
- Using real-time information from the camera, microphone, and screen
- Continuing tasks in the background
- Using AI models and computing resources installed on the device
The EU is not merely requiring Google to disclose the names of these functions. Rival AI services must be able to operate at a level substantially comparable to Google’s own services.
2. Sharing Google Search Data
Google must provide anonymized search data to rival search engines so they can develop and improve their own search services.
AI chatbots that provide search functionality may also access the data if they meet the eligibility requirements.
The shared data may include search queries, search result rankings, web addresses shown to users, and information about clicks and other interactions with search results.
However, Google is not required to disclose its search algorithms or search technology. The purpose is to provide part of the data that competitors need to improve their own search systems.
What Will Rival AI Assistants Be Able to Do on Android?
Android users can currently install AI services other than Gemini, including ChatGPT and other assistants.
However, being able to install an AI app is not the same as having access to the operating system’s major functions.
For an AI assistant to perform tasks on behalf of a user, it needs capabilities such as:
- Voice activation even when the screen is off
- Understanding the current screen and the user’s context
- Opening and controlling other apps
- Writing messages and adding calendar events
- Automating tasks that involve multiple steps
- Continuing to work in the background while the user uses another app
For example, when a user asks a rival AI assistant to write an email, the assistant could open the user’s selected email app, create a draft, and send it.
When asked to order food, it could open a delivery app and complete several steps, including selecting a menu item and entering delivery information.
Rival AI services may also be able to suggest appropriate replies based on a chat conversation or use calendar and location information to provide relevant information in advance.
These capabilities are necessary for rival AI services to move beyond answering questions and develop into AI agents that perform real tasks on smartphones.
When Will Android Features Be Opened?
Google must apply most of the measures to the next major version of Android, Android 18.
The final deadline set by the European Commission is August 1, 2027. The timing can therefore be described as mid-2027, around July or August.
However, the function that allows multiple AI services to recognize their own voice activation phrases at the same time is an exception.
This simultaneous hotword detection feature must be introduced with Android 19, with a final deadline of August 1, 2028.
This means that a system allowing users to activate their chosen rival AI assistant by voice and use major Android functions is expected to begin taking shape in mid-2027. The ability to use multiple AI assistants through different activation phrases may take longer.
Google Search Data Will Also Be Shared with Rivals
The competitiveness of a search service is not determined by its search algorithm alone.
The more data a company collects about what users search for, which results they select, and whether those results were useful, the easier it becomes to improve search quality.
Google has accumulated a large volume of search data over many years. Even when a new search provider develops strong technology, it is difficult to close this data gap quickly.
The EU concluded that this data gap could act as a structural barrier that prevents competitors from entering the market and improving the quality of their search services.
Google must begin preparing an application process, standard agreements, test data, and anonymized datasets during the second half of 2026.
Within six months of the decision, by January 2027, Google must finalize its proposed pricing for data access and provide it to the European Commission and rival service providers.
The shared data will be subject to anonymization measures designed to protect personal information.
User account information and individual search histories will not be provided. Exact search times, unusually long or rare search queries, and precise location information will either be removed or made less specific.
However, not every AI company will be allowed to receive the data.
The recipient must be an eligible provider that actually offers a search service and meets privacy and cybersecurity requirements. The data must be used to develop and optimize search services. Its use for training general-purpose AI models, advertising, or user profiling will be restricted.
Why Google Is Pushing Back
Google argues that opening operating system functions and search data to rival AI and search providers could weaken privacy and cybersecurity.
When an AI agent can access sensitive information such as the camera, microphone, screen, location, and app data, and can also control other apps, the potential impact of a security problem becomes greater.
Search data may also contain sensitive information related to health, finances, work, or political interests. Even when the data is anonymized, Google argues that the risk of identifying users by combining it with other information cannot be completely eliminated.
The EU has required safeguards for user consent, privacy protection, device integrity, and security.
Objective and nondiscriminatory qualification procedures may be applied to certain sensitive functions. Independent third parties will also be able to verify whether rival AI services meet the necessary security requirements.
Is This Decision a Fine?
These measures do not impose a new fine on Google or formally determine that Google has violated the Digital Markets Act.
More precisely, they are binding specification decisions that define how Google must technically implement its existing obligations under the DMA.
A specification proceeding formalizes the regulatory discussions between a gatekeeper company and the European Commission. It establishes how legal obligations should be applied to actual services and technical systems.
The decision itself therefore does not immediately result in a fine.
However, if Google does not implement the required measures, the European Commission may open a separate non-compliance investigation and later determine whether penalties or fines are appropriate.
Google has strongly opposed the decision, arguing that it could weaken privacy and cybersecurity. As of July 19, 2026, the company had not officially announced whether it would take legal action, but it may still challenge the decision before an EU court.
The technical implementation process and the possibility of a legal dispute will therefore need to be closely monitored.
Why This Matters
AI competition is no longer determined only by the quality of a model’s answers.
Even a highly capable AI model faces major limitations in actual use if it cannot be activated by voice on a smartphone, understand the current screen, or open other apps and perform tasks.
By contrast, an AI service operated by the company that controls the smartphone operating system can have several structural advantages:
- It can be activated by voice without opening a separate app.
- It can use information from the screen, location, and apps.
- It can connect multiple apps and perform tasks on behalf of the user.
- It can continue working in the background.
- It can receive new features first through operating system updates.
Platform power in the AI era may therefore depend not only on which company develops the most capable AI model, but also on which AI service is most deeply connected to the operating system’s default functions.
The EU has previously regulated platform features such as app stores, search engines, and payment systems.
This decision is significant because the scope of platform regulation is now expanding to include AI agents’ access to operating systems and data.
DANA NOTES Commentary
This decision does not require Google to remove Gemini from Android or designate a rival AI service as the default assistant.
The key issue is that Google should not use its position as the owner of the operating system to give Gemini exclusive access to major functions while restricting access for rival AI services.
An AI assistant is different from an ordinary app that operates independently on a smartphone.
To function as a true AI agent, it must connect operating system-level capabilities such as voice activation, message writing, calendar access, screen understanding, app control, reservations, and ordering.
A company that controls access to the operating system can therefore secure a powerful advantage in AI competition.
The EU is attempting to distinguish between an advantage created by AI technology and service quality and an advantage created simply by owning the operating system.
The search data-sharing decision follows the same logic.
Google’s accumulated search data is not merely a historical record. It is a core asset used to improve search result quality and the accuracy of AI search services. When rival companies cannot obtain comparable data, it becomes difficult to compete with Google through model quality or search technology alone.
However, this decision is not a fine imposed on Google for unlawful conduct. It is a binding specification decision that clarifies how Google must implement its existing DMA obligations in its Android and search services.
Google has strongly opposed the decision, arguing that it could weaken privacy and cybersecurity. As of July 19, 2026, the company had not officially announced whether it would take legal action, but it may still challenge the decision before an EU court. The technical implementation process and the possibility of a legal dispute will therefore need to be closely monitored.
Ultimately, this decision shows that the unit of competition in the AI market is expanding from models to operating systems and data.
DANA NOTES in One Sentence
The next battleground in AI competition is not only model performance, but also access to operating systems and data.
