Google’s EU data and Android decisions: scope, safeguards and evidence

The European dispute over Google's search data and Android access concerns two different systems. One involves supplying information to eligible competing search services. The other involves access to operating-system capabilities for rival AI assistants. Treating both as an ordinary choice of browser obscures the privacy and engineering questions in each.

There is also a material update to the June 2026 reporting: the European Commission adopted binding specification measures on July 16. An archive explanation should no longer describe the entire decision as something expected the following month. A decision, implementation and evidence of its effects nevertheless remain separate milestones.

What Google warned about in June

WIRED's June 29 report attributed concerns to Google's senior security and privacy staff about re-identifying search users and exposing Android users to fraud. It described Google's claim that internal testing had linked anonymized data back to individuals, while noting that the specific test details had not been published.

The report also recorded disagreement from competitors and researchers about the adequacy of proposed safeguards. Those positions do not establish a measured post-implementation breach rate. A forecast of increased fraud is a forecast; it cannot be rewritten as proof that the increase occurred.

Google has a commercial interest in these obligations, but that does not logically disprove a technical objection. Equally, a security warning by a large platform does not independently prove that a particular implementation is unsafe. The evidence needs to address the actual dataset, controls and capabilities at issue.

July brought two binding sets of measures

The Commission's July 16 announcement identifies separate specifications for Android AI interoperability and search-data sharing. For Android, it describes voice activation of alternative assistants and actions within apps, alongside privacy, device-integrity and security safeguards.

That is more specific than a general claim about opening every app store or changing every default. The search decision concerns competing search services, including AI chatbots offering search functions. It does not mean every company can obtain every category of Google account data for any purpose.

The announcement describes what the measures require and intend to enable. It is not a device-by-device demonstration that every alternative assistant already has identical behavior, or that every permitted interaction has been independently security-tested.

Search access has a defined purpose and recipient boundary

The Commission's search-data explanation identifies ranking, query, click and view information associated with free and paid search, subject to anonymization. It limits use to developing and improving search services, excluding general-purpose model training, unrelated profiling or advertising uses and systematic replication of Google's results.

The same explanation describes technical alteration of data, restrictions on linkage and onward sharing, and independent verification before access with subsequent audits. It also distinguishes information about the person searching from personal information about someone named in a query. Anonymizing the former does not automatically remove the latter.

These details correct the earlier article's speculation that the measure mainly concerned aggregate trends or advertiser-performance dashboards. The scope must come from the identified decision rather than a generic description of what a competition authority might usually mandate.

Anonymity and security answer related but different questions

An anonymization claim concerns whether a person can be identified from the available information in the relevant context. An access-control claim concerns who can obtain information. A dataset could have restricted access and still contain identifying patterns; conversely, removing a direct identifier does not prove every recipient can safely ignore confidentiality.

Contracts and audits address conduct and implementation. They are not substitutes for explaining what technical transformations do. Technical transformations also do not make contractual limits meaningless. Evaluating the combined design requires evidence of both kinds, rather than an assertion that either one alone settles the debate.

This article does not independently reproduce Google's tests or certify the Commission's anonymization method. The public-interest question is whether stated protections hold for the actual data and realistic recipient behavior, including information a recipient can combine with it.

Android access is not the same as receiving search records

An assistant's ability to act in another app raises questions about authority, context and user control. Those differ from questions about a supplied search dataset's identifiability. A finding about one cannot simply be transferred to the other because both involve Google and the DMA.

The earlier article's long VPN, DNS, store-selection and enterprise-configuration checklist did not establish either system's safety. It is removed, together with blanket assurances that a default change resolves the issue. The corrected account records the June concerns, the July decisions and the evidence still needed to assess implementation, without inventing a completed breach, a universal consent rule or a guarantee of harmlessness.