Can gatekeepers use data from multiple services to train AI models?

Not without explicit user consent. Article 5(2) prohibits combining personal data across core platform services for purposes including AI training, unless the user provides specific consent under GDPR Article 6(1)(a).

Does the DMA require portability of AI-processed user data?

Yes. Article 6(9) requires continuous and real-time portability of data generated through user activity, which includes data that AI systems have processed, categorized, or enriched, going beyond GDPR Article 20 in scope and access requirements.

What penalties can the DMA impose for AI-related violations?

The Commission can impose fines up to 10% of the gatekeeper's worldwide annual turnover, increasing to 20% for repeat infringements. Periodic penalty payments of up to 5% of average daily turnover can also be imposed.

Does the DMA apply to AI assistants bundled with operating systems?

Yes. Articles 5(7) and 5(8) restrict tying and bundling practices, preventing gatekeepers from requiring users to use a specific AI assistant as a condition of using the operating system or browser. Choice screens and easy default changes must be provided.

Quick answer

The Digital Markets Act (Regulation 2022/1925) imposes obligations on designated gatekeepers operating core platform services that often rely heavily on AI, including prohibitions on self-preferencing in AI-powered rankings (Article 6(5)), data portability requirements for AI-processed user data (Article 6(9)), interoperability obligations for messaging services (Article 7), and restrictions on combining personal data across services for AI model training without consent (Article 5(2)).

Updated June 2026 · MmowW AI Compliance

Digital Markets Act (DMA) and AI: Gatekeeper Obligations for Algorithmic Services

Q: How does the DMA affect AI-powered search and recommendation algorithms?

Article 6(5) prohibits gatekeepers from self-preferencing in rankings, requiring AI ranking algorithms to apply fair, reasonable, and non-discriminatory criteria. Gatekeepers must document that their AI systems do not favor their own services over competitors.

DMA Gatekeeper Designation and AI Services

The DMA applies to undertakings designated as gatekeepers by the European Commission. A gatekeeper operates a core platform service (CPS) with significant impact on the internal market, serves as an important gateway for business users to reach end users, and holds an entrenched and durable position. As of 2026, designated gatekeepers include Alphabet (Google Search, Maps, Android, YouTube, Chrome, Google Ads, Google Shopping, Google Play), Amazon (Marketplace, Ads), Apple (iOS, App Store, Safari), ByteDance (TikTok), Meta (Facebook, Instagram, WhatsApp, Messenger, Meta Marketplace, Meta Ads), and Microsoft (Windows, LinkedIn). Nearly all designated CPSs rely on AI systems for core functionality: search ranking, content recommendation, advertising targeting, and marketplace matching.

Self-Preferencing Prohibition and AI Rankings

Article 6(5) prohibits gatekeepers from treating their own services more favorably in ranking than similar third-party services. For AI-powered ranking algorithms (search results, marketplace product listings, app store recommendations), this requires the ranking system to apply fair, reasonable, and non-discriminatory criteria without giving preferential treatment to the gatekeeper's own offerings. The European Commission's investigation into Google's compliance (2024-2025) examines whether its AI-powered search ranking complies with this obligation, building on the General Court's confirmation of the Google Shopping abuse of dominance finding (T-612/17).

DMA Obligation	Article	AI Impact	Gatekeeper Action Required
No self-preferencing in ranking	Art. 6(5)	AI ranking algorithms must be neutral	Audit ranking systems, document criteria
Data portability	Art. 6(9)	Export AI-processed user data	Provide tools for continuous data export
Data combination restriction	Art. 5(2)	Cannot combine data across CPSs for AI training	Separate data pipelines per service
Interoperability	Art. 7	Messaging AI features must interoperate	Open protocols for basic messaging
Ranking transparency	Art. 6(5) + P2B Reg.	Disclose AI ranking parameters	Publish ranking methodology descriptions
No tying of AI services	Art. 5(7)-(8)	Cannot force AI assistant with OS/browser	Allow default changes, offer choice screens

Data Combination Restrictions and AI Training

Article 5(2) prohibits gatekeepers from combining personal data from their CPS with data from other services offered by the gatekeeper or third parties, or cross-using personal data from one CPS in other services, unless the end user has given specific consent under GDPR Article 6(1)(a). This has profound implications for AI model training: a gatekeeper cannot combine user data from its search engine with data from its email service to train a unified AI model without explicit user consent. Given that most users do not consent when given a genuine choice, this effectively fragments the data available for AI training across the gatekeeper's services.

Data Portability for AI-Processed Data

Article 6(9) requires gatekeepers to provide end users and authorized third parties with effective portability of data generated through the user's activity on the CPS, including continuous and real-time access. For AI-driven platforms, this includes data that AI systems have processed, categorized, or enriched based on user activity: recommendation preferences, search history analysis, behavioral profiles, and content interaction patterns. This obligation goes beyond GDPR data portability (Article 20) in scope and real-time access requirements, enabling competitors to build AI services on data that users choose to export.

Interoperability and AI-Powered Messaging

Article 7 requires gatekeepers providing number-independent interpersonal communication services (messaging) to provide interoperability with third-party services upon request. As messaging platforms increasingly integrate AI features (smart replies, content summarization, translation), the interoperability obligation raises questions about whether AI-enhanced features must also interoperate. The current obligation covers basic messaging functionality (text, images, voice messages, video calls, group chats in phased implementation), but AI features layered on top are within the scope when they affect the basic interoperable message delivery.

Compliance Framework for AI Components

Gatekeepers must submit a compliance report within six months of designation, detailing measures taken for each CPS. For AI-powered services, this requires: (1) documenting how ranking algorithms satisfy the non-discrimination requirement of Article 6(5); (2) demonstrating data separation measures that prevent unauthorized cross-service data combination for AI training; (3) implementing data portability infrastructure that covers AI-processed user data; (4) providing ranking transparency descriptions compliant with both the DMA and the P2B Regulation (2019/1150); and (5) ensuring AI assistant services are not unfairly tied to operating systems or browsers. The Commission can impose fines up to 10% of worldwide turnover, rising to 20% for repeat infringements.

Check your AI compliance readiness — free.

Take the Readiness Check 3 minutes · 10 questions · no signup required

This article is for informational purposes only and does not constitute legal advice. Regulatory requirements change frequently — verify current rules with official sources. Built by Sawai Gyoseishoshi Office, Hiroshima, Japan.