On July 16 the European Commission handed Google two binding specification measures under the Digital Markets Act. One requires that rival AI assistants — such as OpenAI’s ChatGPT and Anthropic’s Claude — get the same system-level Android access as Google’s own Gemini. The other requires Google to share the data it has built up through Search with competing search engines and AI chatbots. As it happens, today, July 27, was the legal deadline for the Commission to finalize the decision — and it had already adopted it 11 days earlier, on the 16th.

TL;DR 📝

  • The EU imposed two binding decisions on Google: ①Android AI interoperability and ②search data sharing (adopted July 16; legal deadline July 27).
  • Users will be able to summon ChatGPT or Claude by voice, like “Hey Google,” and have them act inside apps (covering roughly 60% of EU users, who are on Android).
  • Non-compliance carries DMA fines of up to 10% of global revenue (20% for repeat offenses). That said, actual rollout is slated for 2027, so it will take time.

📌 What exactly did the EU order Google to do?

The decision comes in two distinct strands. The Commission described both as “binding specification measures” grounded in the Digital Markets Act (DMA).

One strand is Android AI interoperability. Today, on Android phones, rival AI assistants can only reach the operating system’s core functions in a limited way, while Google’s Gemini roams across the entire system. The Commission concluded that this asymmetry distorts competition. The other strand is Google search data sharing. The goal is to level a tilted playing field by making Google share the search data only it can gather at scale with rival search engines.

🎙️ Can you drive your phone with ChatGPT or Claude the way you do with “Hey Google”?

According to the Commission, users will soon be able to activate the AI assistant of their choice by voice command — much like calling up Gemini with “Hey Google” today.

The examples the Commission gives are fairly concrete. A user could ask a third-party AI assistant to handle a task like booking a taxi, receive suggested replies in a chat app, or ask about a place they recently visited. In short, rivals would be allowed to offer the same kind of “agent” functions that move across apps and act on the user’s behalf. The Commission stressed that even as it opens these doors, it has built in safeguards to protect user privacy and the integrity and security of devices. The scope is not small: the Commission says about 60% of EU users are on Android, and some outlets report that Google holds roughly 65% of Europe’s mobile operating system market.

🔍 Why is search data sharing the crux of the matter?

The second decision spells out in detail how Google must share its search data, on the view that data sharing is decisive for the development and optimization of third-party search engines.

The Commission noted that Google’s earlier data-sharing offers had been ineffective, and it nailed down several points. AI chatbots that offer search functions are also eligible to receive the data, and Google must share the same data it collects to optimize its own search (subject to anonymization). The anonymization method was designed in multiple layers together with internal and external privacy experts, aligned with the draft joint guidelines on the interplay between the DMA and the GDPR. The measures also include a clause letting Google first assess whether sharing with a specific party poses serious cybersecurity or data-protection risks, a fair formula for pricing the shared data, and a transparent access process.

🗓️ When does anything actually change?

The decision is out, but it will take time before the changes reach users. This is not a sudden move — it is the conclusion of a process that has run for more than half a year.

The Commission opened the two specification proceedings on January 27 this year, published preliminary findings on the search data on April 16, and adopted the final decision on July 16. Earlier reporting had put the final deadline at July 27 — today — so the Commission beat it by more than ten days. As for actual implementation, some outlets report that search data access starts in January 2027, that most Android AI features roll out in stages through August 2027, and that some features extend into 2028. The detailed schedule and the number of feature items, however, are not spelled out as such in the Commission’s official announcement, so the exact timing needs further confirmation.

⚖️ Why is Google pushing back, and what happens if it doesn’t comply?

Google has voiced concern that the measures could undermine user privacy, security, and innovation. “Android is open by design” is the company’s line.

Google points to the fact that users can already download any AI app from the Play Store. The Commission’s response is that being downloadable and having equal access are two different things. An AI assistant that cannot be summoned by voice, cannot read the screen, and cannot connect to the operating system’s core apps is hard to call a genuine competitor, regardless of whether it can be downloaded. The enforcement teeth are considerable: the DMA allows fines of up to 10% of global annual revenue, rising to 20% for repeat violations. Some outlets estimate that 10% of Alphabet’s revenue would top $30 billion. And because a DMA appeal does not automatically suspend the obligation to comply, Google must implement the measures on schedule even while contesting them in court.

🍎 What does this mean for Apple and for Korea?

This is not just Google’s problem — it reflects a broader shift across smartphone platforms: the pressure to treat AI assistants as a “contestable layer” rather than a proprietary feature.

Apple has faced the same pressure. According to some outlets, Apple has let EU users set a voice assistant other than Siri as the default in iOS 26.2, and there is talk of Apple opening Siri to AI services beyond its existing ChatGPT partnership in iOS 27. Still, it is worth being clear that this remains an EU-only regulation for now; it does not apply directly to Korea or other regions. Even so, for domestic users and the AI and platform industry, it is worth watching as a precedent for how regulators handle a structure in which the owner of an operating system embeds its own AI deep into the system.

Overall take 📊

The message of this EU decision is clear: it aims to stop the AI-assistant race from being decided by “who controls the operating system” rather than “whose model is smarter.” If a few days ago we looked at frontier AI models competing on performance and price, this case shows the platform-layer fight beneath it — over who gets to sit in the phone’s default-assistant seat.

Three things are worth watching. First, the decision is out, but actual implementation is slated for 2027, so it will take time for users to feel the change. Second, since Google’s obligation to comply does not pause even if it appeals, legal battles and real-world rollout are likely to proceed side by side. Third, this is an EU-only rule for now, but with Apple moving in a similar direction, whether the principle spreads to other regions is the key thing to watch. Details such as the exact rollout schedule and the feature items that are not stated in the official announcement will need to be confirmed as follow-up material emerges.

※ This article is for informational purposes only and is not investment advice.

Sources 🔗