Summary

• The European Commission has imposed binding Android interoperability and search data requirements on Google under the Digital Markets Act.
• Competing AI assistants are due to gain access to Android functions, while eligible search and AI services will receive anonymised Google search data.
• Commercial effects will depend on latency, privacy controls, technical documentation, pricing, and whether access proves usable in production.

The European Commission has set binding technical conditions for how Google must open Android and parts of its search data to competitors, moving enforcement under the Digital Markets Act from general market obligations into the machinery of mobile devices and search services.

Two specification decisions require Google to give competing artificial intelligence assistants access to Android functions and provide eligible search providers with anonymised ranking, query, click, and viewing information. Brussels wants to prevent Gemini from receiving preferential access through Google’s control of Android, while reducing the data advantage held by a search engine used at enormous scale.

Rather than ordering a corporate break-up or imposing another financial penalty, the Commission is attempting to change the conditions under which rival products operate. An AI assistant cannot become a credible replacement for the default service when it can answer questions but cannot reliably invoke device functions, understand on-screen context, or complete actions across applications.

The Android measures cover 11 functions and follow a specification process under Article 6(7) of the DMA. Google must provide access on terms that are effective, proportionate, and non-discriminatory, while the Commission will monitor implementation for two years. Developers can consult the Commission’s Android interoperability material as they assess whether the remedy supports workable services.

Search data creates a different engineering and commercial problem because the information has to be useful without exposing personal data, trade secrets, or systems that could be manipulated. Eligible search engines, including some AI services with search functions, should be able to obtain anonymised information covering queries, rankings, clicks, and views.

Google’s earlier proposal failed to satisfy the Commission because its anonymisation approach removed between 90% and 100% of unique queries from the dataset, leaving rivals with information that offered limited competitive value. The final specification requires a maximum latency of seven days and allows successful applicants to retain access for up to five years.

Access still has to survive production use

Detailed interoperability rules can alter product development, although access alone will not erase Google’s advantages in distribution, brand recognition, infrastructure, developer relationships, and accumulated behavioural information. Smaller AI and search companies still have to fund services that work across languages, devices, and regulatory regimes while meeting the reliability expected of software embedded in daily work and communications.

Developers that can invoke system functions, obtain relevant context, and act across applications can build deeper workflows than those confined to a standalone chat window. Enterprise software suppliers may begin to treat Android as a more open automation layer, but every additional connection raises questions about identity, permissions, data protection, and the audit records created when an assistant acts for an employee.

Search data could have an equally broad effect because retrieval remains an essential component of many generative AI products. Training a language model does not create a dependable, current search engine. Query patterns, click behaviour, ranking feedback, and freshness signals help providers understand what people seek and whether results are useful, making the underlying data commercially valuable even after anonymisation.

Formal availability can still conceal practical exclusion when interfaces are expensive, incomplete, delayed, poorly documented, or changed at short notice. Earlier platform remedies have left incumbents controlling test environments, documentation, release schedules, and technical support, allowing access to exist in legal terms while remaining unattractive to most developers.

By specifying latency, eligibility, data categories, and Android functions, the Commission is trying to close some of those routes. Enforcement will nevertheless require close examination of uptime, response times, parity with Google’s own services, contractual conditions, and the cost of adapting products whenever Android changes.

Privacy and security will remain legitimate constraints rather than inconveniences to be waved away in the name of competition. An assistant with extensive device permissions can reach messages, business systems, location data, and authentication processes. Google may argue that some requests create unacceptable exposure, while rivals will challenge safety claims that conveniently preserve the incumbent’s control.

Brussels will have to distinguish between genuine technical risk and safeguards designed so broadly that they prevent competition. It must also avoid compelling access that weakens the security model of devices used across banking, healthcare, public services, and corporate networks.

The DMA is entering a phase in which its credibility will depend less on the number of proceedings opened than on the markets created after remedies take effect. Android and search now provide a direct test: either developers gain enough dependable access to build credible alternatives, or one of Europe’s most ambitious competition laws produces interfaces that are open on paper and commercially quiet in practice.