22SEP

DMA orders Google to open search data

4 min read

10:47UTC

Brussels sent Google preliminary DMA measures requiring it to share search rankings, queries and click data with rivals and AI chatbots on FRAND terms. A binding decision is due by 27 July.

← Back to Germany's AI champion merges into Cohere Jump to analysis ↓

TechnologyDeveloping

Key takeaway

Google must share search data with rival chatbots, an involuntary data dividend for European AI.

DG COMP and DG CNECT, the European Commission's competition and digital-strategy directorates, sent Google preliminary compliance measures on 16 April 2026 requiring it to share search rankings, user queries, click data and view metrics with rival search engines and AI chatbots on fair, reasonable and non-discriminatory (FRAND) terms ¹. A public consultation opened on 17 April and closes on 1 May; a binding decision is due by 27 July 2026 ¹. Search rankings, queries and click logs are the foundational training substrate every frontier AI assistant needs; the order is an involuntary data dividend for European AI builders who cannot scrape at Google scale.

Executive Vice-President Teresa Ribera framed the order around AI alongside search: "Data is a key input for online search and for developing new services, including AI" ¹. The technical annex sets eligibility criteria, delivery frequency, anonymisation procedures and governance ¹. Google responded that the demands threaten user privacy on "sensitive searches about health, family, finances" and alleged the investigation was "driven at least in part by OpenAI" ¹. Google is asking Europe to read the remedy as a US-on-US proxy fight rather than a European sovereignty instrument.

The order extends the DMA lineage from platform fines to a behavioural data-access remedy. European firms that cannot scrape at Google scale, Mistral AI included, would gain an involuntary data dividend if FRAND terms survive the consultation. The 27 July binding decision deadline sits five days before the AI Office, the Commission's AI enforcement body, gains full enforcement powers over general-purpose AI model providers on 2 August . That creates a regulatory collision between competition and AI enforcement arms with no public joint pipeline connecting the two.

Deep Analysis

In plain English

When you search on Google, the search engine records what you typed, which results you clicked on, and which ones you ignored. Over 25 years, Google has collected billions of these records. They tell an AI system what information people actually want when they type different queries, giving it 25 years of annotated signal that no AI company can buy or scrape from scratch at equivalent scale. The EU's Digital Markets Act (DMA) is a law that says very large tech platforms must play by special rules to keep markets fair. The European Commission has told Google it must share this search data with rival search engines and AI chatbot makers, on terms that are fair and non-discriminatory. Google argues this threatens user privacy. The Commission says it is essential for European companies to be able to build AI systems that can compete with US ones. A formal decision is due on 27 July 2026.

Deep Analysis

Root Causes

European AI companies face a structural training-data deficit. Google processes roughly 8.5 billion searches per day and has accumulated two decades of annotated query-click feedback loops. Mistral AI, founded in 2023, cannot replicate that dataset through commercial scraping or licensed data partnerships at equivalent quality or volume.

The AI Act's forthcoming general-purpose AI transparency requirements will require model providers to document training data provenance and demonstrate diversity; mandated access to Google search logs addresses both the data-volume gap and the provenance documentation problem in a single remedy.

The Draghi report on European competitiveness identified data access as one of the three structural barriers to European AI parity alongside compute and talent. The DMA order on search data is the Commission's first concrete implementation of that Draghi recommendation, though it targets one specific platform's dataset rather than establishing a general European data-access framework.

What could happen next?

Consequence
The 1 May consultation close and 27 July binding decision creates a 12-week implementation clock; Google's legal challenge will likely begin before the ink is dry, extending effective implementation by 12 to 18 months.
Short term · 0.78
Precedent
A sustained DMA search-data remedy would establish the principle that gatekeeper platform data constitutes quasi-public infrastructure, with implications for Meta social graph data, Apple App Store behavioural data, and Amazon marketplace data in subsequent proceedings.
Medium term · 0.72
Opportunity
The binding decision landing five days before the AI Act enforcement date on 2 August creates a regulatory stack where European AI providers gain both a data-access right and an enforcement environment that disadvantages US competitors simultaneously.
Medium term · 0.65

Source Landscape

This story draws on neutral-leaning sources from Belgium

Belgium

Primary parallel: The 1980s US-Japan Semiconductor Trade Agreement of 1986 required Japan to open its domestic semiconductor market to foreign suppliers and share process technology under bilateral government pressure.

The forced licensing created a temporary competitive floor for US firms, but Japan's chipmakers adapted faster than the agreement expected, and by the mid-1990s Korean manufacturers had used the same forced-technology environment to overtake both. The DMA search-data order follows the same structural logic: mandatory data sharing as a catchup mechanism for laggard competitors.

Counter-parallel: The European Commission's 2004 Microsoft interoperability remedy required Microsoft to share server communications protocols with rival software vendors on FRAND terms. After a decade of legal proceedings and technical disputes over pricing and format, the remedy produced limited practical market entry.

The primary beneficiary was Samba, an open-source project, rather than commercial European software challengers. If FRAND search data sharing follows a similar trajectory, the beneficiary may be open-source AI projects rather than the Mistral AIs and Aleph Alphas the Commission intends to advantage.

Google's search advertising revenue in Europe totalled approximately €28bn in 2024. The query and click-data stream that produces that revenue contains implicit relevance judgements and user intent signals accumulated over 25 years of product iteration. Analysts at Gartner estimate that high-quality search log datasets of this kind would cost $500m to $2bn to replicate from scratch for a foundation model training programme.

For European AI companies, mandatory FRAND access at a commercially reasonable price converts a capital acquisition problem into an operating expense. Mistral AI's $830m debt raise was allocated almost entirely to GPU compute; access to Google-quality search training data would not require additional capital expenditure, only a licensing fee. The value transfer is from Google's data moat to European AI training budgets, at a price the Commission's technical annex will determine.

Consensus view: The European AI Forum and Bruegel senior fellow Georgios Petropoulos argue the FRAND data-sharing order creates a genuine competitive substrate for the first time, provided the technical annex sets enforceable delivery frequencies and anonymisation standards that prevent Google from fulfilling the letter of the remedy while degrading data quality below usefulness.

The binding decision deadline of 27 July 2026 gives the Commission five days before the AI Office gains full enforcement powers on 2 August, creating an intentional regulatory handoff: DMA competition remedy feeding into AI Act compliance stack.

Counter-view: The Computer and Communications Industry Association (CCIA), which represents Google, argues that FRAND search data sharing is a European solution to a global problem that US competitors created. Google's public statement frames the order as an OpenAI-lobbied proxy intervention rather than a structural competition remedy.

CCIA research economist Christian Reimsbach-Kounatze has argued that mandated data sharing creates free-rider dynamics where smaller firms extract value from Google's continuous quality investment without bearing the infrastructure cost, ultimately degrading the shared resource.

Key tension: Whether the anonymisation requirements that protect user privacy on sensitive queries (health, finance, family) can be calibrated to preserve enough signal for AI training, or whether privacy-legitimate redactions will strip precisely the query categories that carry the highest signal density for training contextually aware European language models.

Sources:European Commission DG CNECT·European Commission DG CNECT

Mentions:DG COMP →DG CNECT →Google →OpenAI →Mistral AI →AI Office →Microsoft →Japan →Aleph Alpha →European Commission →Bruegel →Digital Markets Act →Teresa Ribera →

First Reported In

Update #2 · Brussels buys, Britain backs, Google unlocks

European Commission DG CNECT· 19 Apr 2026

Read original →

Causes and effects

Caused by

Brussels fines Apple, Meta, and X €820m

The DMA Google search-data preliminary measures extend the DMA lineage from platform fines to a behavioural data-access remedy.

Occurred 1 Jan 2026

Read story →

AI Office gains enforcement powers in August

The 27 July DMA Google binding decision deadline sits five days before the 2 August AI Act full-enforcement date, creating a regulatory collision point.

Occurred 2 Aug 2026

Read story →

This Event

DMA orders Google to open search data

The order extends DMA enforcement from headline fines to behavioural data access, handing European AI builders training substrate their scraping budgets cannot buy.

Led to

DMA clock starts toward six-day AI Act collision

The DMA.100209 consultation opened in this event is the public phase of the preliminary measures Google received in ID:2619; the consultation formalises the process toward the binding 27 July decision.

Occurred 16 Apr 2026

Read story →

Google fine stuck on von der Leyen's desk

The 27 July Google search-data sharing ruling creates the double-barrel enforcement context that makes the fine delay more consequential.

Occurred 1 May 2026

Read story →

Different Perspectives

ESMC (TSMC-majority joint venture)

ESMC's president said construction remains on schedule after the Dresden fab's topping-out ceremony on 14 September, reported by Focus Taiwan with first process equipment still targeted for the second half of 2027. No first-party ESMC or TSMC statement independently confirms the claim, and the fab remains 70% TSMC-owned inside a project Europe cites as its semiconductor sovereignty case.

Civo

Civo sold out its Navigate London sovereignty conference on 22 September, drawing about 800 attendees including a sitting MP, a former defence procurement minister and sponsors led by Nokia. Companies House confirms chief executive Mark Boost as Civo's sole person with significant control, British and UK-resident, which answers the ownership question the conference itself is arguing matters.

United States Trade Representative

USTR opened its 2027 National Trade Estimate comment window on 14 September, naming the EU among markets with restrictive technology requirements and inviting submissions on cross-border data rules. The window follows Trump's 24 July Section 301 order into EU digital rules by seven weeks, and unused comments are kept, in USTR's own wording, for future negotiations.

Cohere

Cohere published the deal on 16 September without naming a regulator, running the merged company globally under its own brand from dual Toronto and Berlin headquarters. It pledges the combined company will deliver sovereign AI on STACKIT, the Schwarz Group's German platform, aimed at government buyers weighing that offer against Berlin's own anchor-customer signal.

Germany (Federal Government)

Digital Minister Karsten Wildberger called the Cohere talks "a very strong signal" and signalled Berlin's readiness to become an anchor customer, now its main lever since equity sits with Cohere. The German side secured a co-headquarters and two Cohere C-suite seats, but the protective-rights terms it pressed for in July remain undisclosed.

Poland

Poland leads a self-announced AI Gigafactory consortium with a EUR 100 million phase-one commitment, matched by Czechia and joined by Hungary at EUR 25 million. EuroHPC has confirmed no consortium for the call closing 12 November, so the bloc exists only in national announcements so far.