The European Commission opened two parallel DMA specification proceedings against Google on January 27, 2026 — the first DMA specification proceedings targeting AI assistant access on a major mobile platform. Announced by Henna Virkkunen (Executive Vice-President for Tech Sovereignty, Security and Democracy) and Teresa Ribera (Executive Vice-President for a Clean, Just and Competitive Transition), the proceedings must conclude within six months. If Google fails to comply with the resulting measures, it faces fines up to 10% of global revenue — roughly $35 billion based on Alphabet’s 2024 revenue of $350 billion. The target isn’t Google’s AI technology. It’s the distribution advantage baked into Android and Search that locks competitors out before they can even compete. The European Commission’s announcement frames this as assistance toward compliance, but make no mistake — this is enforcement with a timer.
The two proceedings: Android AI access and search data sharing
The first proceeding targets Android AI interoperability under Article 6(7) of the DMA. Google must provide “free and effective” access to the same Android system features that Gemini enjoys exclusively. That means on-device AI processing, system-level APIs for screen reading and app control, battery optimization frameworks, access to neural processing units (NPUs), voice activation from any screen, and lock screen integration. Right now, third-party AI assistants like Claude, ChatGPT, and Grok exist as app-only utilities on Android — they can’t do what Gemini does as a deeply integrated system assistant.
The second proceeding forces Google to share anonymized search data under Article 6(11) on fair, reasonable, and non-discriminatory (FRAND) terms. This includes ranking signals, query data, click-through rates, and view data — data from Google’s European Search Dataset Licensing Program, which covers a substantial volume of European search queries. DuckDuckGo has publicly criticized Google’s existing data-sharing efforts as “unusable for search training purposes” due to aggressive anonymization that strips context. The open question: whether AI chatbot providers like Anthropic and OpenAI — not just traditional search engines — qualify for access to this data under the DMA’s definition of “business users.”
The Commission will communicate preliminary findings within three months and must conclude the proceedings within six. As Engadget reports, if Google doesn’t satisfy the resulting requirements, non-compliance fines scale with severity and repetition.
This follows a familiar pattern of EU regulatory pressure
The EU DMA Google Gemini proceedings follow the iOS interoperability specification proceedings Apple faced starting September 2024 — the first use of this DMA mechanism — where the Commission demanded equal third-party access to system features. Google has accumulated €8.25 billion in EU antitrust fines between 2017 and 2019 for Android bundling, search manipulation, and AdSense restrictions. The specification proceeding is the EU’s “carrot before stick” — a compliance assistance framework that becomes an enforcement action if ignored.
The timing is deliberate. The DMA undergoes its first mandatory review in May 2026, just four months away. The EU AI Act becomes fully applicable in August 2026, creating regulatory convergence across three frameworks: DMA, AI Act, and the Digital Services Act (DSA). Google now navigates simultaneous compliance requirements across competition law, content moderation, and AI safety. Android Authority’s analysis notes this is the first time the Commission has explicitly targeted AI distribution mechanisms rather than algorithmic fairness or transparency.
Worth noting: this builds on AI policy discussions at Davos earlier this month, where European regulators signaled aggressive enforcement of existing frameworks before creating new ones.
Google’s response: privacy, security, and innovation concerns
Clare Kelly, Google’s Senior Competition Counsel, responded that “Android is open by design… we are concerned that further rules will compromise user privacy, security and innovation.” The privacy and security concerns are legitimate in isolation — granting system-level access to third-party AI models does introduce new attack surfaces and data-handling complexities. But Google made identical arguments during prior antitrust proceedings about Android app defaults, search placement, and browser choice screens. The privacy argument is simultaneously true and convenient.
The real concern isn’t whether Google can manage privacy-preserving interoperability — they already do this with complex permission systems across thousands of apps. The concern is competition. Gemini’s system-level integration on Android isn’t just a feature advantage; it’s a moat that prevents competitors from offering comparable experiences regardless of model quality.
Who benefits and how much access changes
| AI Assistant | Current Android Status | What Changes |
|---|---|---|
| Claude (Anthropic) | App-only | System-level access, voice activation, sensor integration |
| ChatGPT (OpenAI) | App-only | System-level access, voice activation, sensor integration |
| Grok (xAI) | App-only | System-level access, voice activation, sensor integration |
| Perplexity | App-only | System access + search training data |
| DuckDuckGo | Search app | Direct search training data, voice integration |
Android holds roughly 72% global smartphone market share according to 2025 data. Getting system-level access on Android represents the single biggest distribution unlock available to any AI assistant. Anthropic and OpenAI gain the most — both have frontier models competitive with or superior to Gemini in various benchmarks, but were locked out of the world’s most popular mobile operating system at the infrastructure level.
For search-focused competitors like Perplexity and DuckDuckGo, access to Google’s anonymized search dataset provides training signal that’s been unavailable at scale. This doesn’t eliminate Google’s data advantage — they still capture more queries, more click feedback, and more context than any dataset they’re required to share. But it gives rivals enough signal to train competitive ranking systems and validate new approaches. The proceedings acknowledge what Gemini’s recent capabilities demonstrated: modern AI assistants compete on distribution and integration as much as model intelligence.
The distribution shift: convenience beats capability
The EU just reframed AI competition from “whose model is smartest” to “whose assistant is most accessible.” By dismantling Gemini’s exclusive Android integration, the Commission shifts competition toward model quality and user experience rather than platform lock-in. This is the most consequential structural intervention in AI competition since the EU AI Act passed — and arguably more impactful, since it targets the bottleneck that determines which AI assistants billions of users can actually choose.
The search data proceeding operates on similar logic. Google’s ranking algorithms benefit from decades of query feedback, click-through analysis, and iterative refinement that no competitor can replicate from scratch. Sharing anonymized datasets doesn’t eliminate that advantage, but it provides enough training signal for rivals to build credible alternatives. DuckDuckGo and Perplexity benefit most here — both already have search products but lack the data scale to compete on ranking quality.
The philosophical shift matters. Traditional antitrust focused on preventing monopolistic abuse after dominance was established. The DMA attempts to prevent dominance from forming by mandating interoperability before winner-take-all dynamics lock in. Whether this works depends on execution, but the framework represents a genuine departure from US-style competition enforcement. Worth reading: Anthropic’s constitutional approach to AI development, which prioritizes safety and values alignment — areas where distribution access could actually accelerate adoption of more careful AI systems.
Timeline and what to watch next
The Commission will issue preliminary findings within three months — likely by late April 2026. The Commission must conclude the specification proceedings by approximately July 27, 2026 — six months from opening. Non-compliance with the resulting measures triggers fines up to 10% of global revenue. For repeated non-compliance, fines scale to 20% of revenue. The EU AI Act becomes fully applicable in August 2026, one month after the DMA compliance deadline, creating overlapping regulatory pressure.
What to watch: whether Google appeals the proceedings (likely, given past behavior), whether this inspires parallel actions in the UK or other jurisdictions, and whether Android engineering actually changes or Google finds creative compliance workarounds that satisfy the letter of the law while preserving competitive advantages. Also watch how third-party AI developers respond — whether they invest in Android system integration or wait to see if Google’s compliance is genuine. WebProNews characterizes this as an ultimatum with enforcement teeth, not a negotiation.
The EU just made a structural bet: that AI competition is decided by distribution access, not just model capability. Google’s monopoly on where AI assistants live on the world’s dominant mobile platform is now under direct legal attack. Whether rivals like Anthropic, OpenAI, and xAI can capitalize depends on model quality and user experience — but for the first time, they’ll get to compete on equal infrastructure footing. That’s the theory, anyway. Implementation determines whether the DMA becomes a turning point or a cautionary tale about regulatory overreach. ABC News coverage notes this is being watched closely by US antitrust enforcers, who face similar questions about AI platform dominance but lack equivalent regulatory frameworks.
Get the Daily Pulse
Sharp analysis on what's actually moving in AI. No hype, no filler, no weekly digest.
