December 2025: Two Probes That Changed the AI Regulatory Landscape
The European Commission’s December 2025 double announcement — investigations into both Meta and Google on AI-related competition grounds — marked a turning point in how European authorities approach AI market power. For the first time, antitrust enforcement was being deployed not to address historical market concentration in search or social media, but to examine the real-time dynamics of AI integration into dominant platforms.
According to Hausfeld’s competition law analysis, the Commission’s investigation into Meta centres on WhatsApp’s policy restricting third-party AI assistants from accessing users through the platform. The theory of harm is a classic bundling concern: Meta is accused of using its dominant messaging platform to channel users exclusively toward its own Meta AI chatbot, foreclosing the market for competing AI assistant providers who would otherwise benefit from WhatsApp’s 2+ billion global user base.
The Commission’s Google investigation is structurally different but equally significant. According to WSGR’s 2026 antitrust preview, the probe examines whether Google used publisher web content for its AI Overviews and AI Mode features “without appropriate compensation to publishers and without offering them the possibility to refuse.” A parallel track examines whether YouTube training data was used for AI model development without compensating creators or allowing opt-out.
Why the DMA Isn’t Doing This Work — and Why That Matters
The most analytically significant aspect of both investigations is the Commission’s choice of legal instrument. Despite Meta and Google both being designated as “gatekeepers” under the Digital Markets Act — a designation that should trigger the most robust form of platform regulation Europe has created — the Commission chose to proceed under Article 102 TFEU, the traditional competition law provision prohibiting abuse of dominant market position.
EC Commissioner Ribera’s public statements, reported by Hausfeld, make the reason explicit: while aspects of Google’s conduct “could have worked” under the DMA, traditional competition law was “much more effective” for enforcement. This is a candid admission that the DMA — designed over 2019-2021, before generative AI became commercially significant — has structural gaps when it comes to novel AI services.
The core gap is definitional. The DMA’s core platform service categories (online search, social networking, messaging, operating systems, browsers, virtual assistants, advertising services, and online intermediation) were written with existing platform categories in mind. AI services like Google’s AI Overviews, Meta’s AI chatbot integration, and standalone large language models are not explicitly listed as core platform services. The Commission is currently reviewing whether to expand the DMA’s scope to include generative AI — a consultation that closed in May 2026 — but the outcome of that review will shape enforcement for years without providing a solution for the cases already open.
The Meta WhatsApp AI Case: Bundling in the AI Era
Italy’s competition authority (AGCM) was first to move on the Meta WhatsApp AI issue, launching an investigation in July 2025 into Meta’s pre-installation of Meta AI on WhatsApp. By November 2025, the investigation expanded to cover Meta’s terms allegedly “excluding competing AI chatbots from using WhatsApp.” In December 2025, the AGCM adopted interim measures prohibiting those restrictive terms in Italy.
The European Commission’s formal investigation, opened in December 2025, covers the same conduct at EU-wide scope (with Italy carved out due to AGCM’s parallel action). By February 2026, the Commission had notified interim measures under Article 102 TFEU — an unusual step that indicates regulators believe consumers are being actively harmed while the investigation proceeds.
The market theory is straightforward: WhatsApp’s near-universal penetration in Europe — over 80% of messaging app users in multiple member states — means that exclusionary terms in WhatsApp’s API access conditions effectively determine which AI assistants can reach the majority of European consumers. An AI assistant that cannot integrate with WhatsApp is structurally disadvantaged relative to Meta AI, which operates as the platform’s default.
For AI companies building assistant and agent products, the case has direct practical implications. If the Commission determines that Meta’s API restrictions are anticompetitive, it could order open-access remedies that would allow third-party AI assistants to integrate with WhatsApp — creating a distribution channel that is currently closed. No AI-specific fines have been announced in either investigation as of May 2026; formal decisions typically take 12 to 24 months from the opening of formal proceedings.
Advertisement
The Google Content Case: Who Owns the Training Data?
The Google investigation raises what may be the defining ownership question of the AI era: when a dominant platform uses third-party content to train AI systems that then compete with the original content creators, is that use lawful under competition law?
ProMarket’s analysis of 2026 European antitrust trends frames the Google case within a broader pattern the Commission is establishing: that AI’s dependency on training data creates a new category of potentially anticompetitive input foreclosure. Publishers and news organisations whose content trains AI systems that generate summaries — reducing traffic to the original source — are in a position analogous to companies whose inputs have been appropriated by a dominant buyer.
The Commission’s January 2026 proceedings to clarify Google’s data-sharing obligations with AI chatbot providers add a second dimension to the case. Beyond the training data question, the Commission is examining whether Google’s privileged access to its own search data and advertising signals gives its AI systems a structural advantage that cannot be replicated by competing AI providers — even those with equal technical capability and compute resources.
The Goodwin Procter year-in-review analysis notes that the Commission’s September 2025 €2.95 billion fine against Google for ad tech dominance abuse established that large fines for platform conduct are a credible enforcement tool — not just a theoretical risk. That context gives the AI-related investigations teeth: Google knows the Commission is willing to impose nine-figure penalties when it finds violations.
What AI Companies and Platform Operators Should Do
1. Audit Your Platform Integration Terms for AI Exclusivity
The Meta WhatsApp case establishes that platform operators with dominant market positions cannot impose API terms that exclude competing AI providers — at least under EU competition law. If your company operates a platform or messaging service with significant market share in any EU member state, your API and integration terms need legal review for AI exclusivity provisions. Even terms that appear commercially reasonable — preferential placement for your own AI, reduced API limits for competitors — can constitute anticompetitive conduct under Article 102 when combined with dominant market position.
For AI companies that are currently excluded from major platform APIs (WhatsApp, iMessage, other dominant messaging services), the Meta case creates a monitoring opportunity: if the Commission orders open access remedies, affected companies should be prepared to apply for API access quickly, as early distribution partners will build user relationships that are difficult for later entrants to displace.
2. Document Your Training Data Provenance and Compensation Arrangements
The Google investigation establishes that training data sourcing is becoming a competition law issue, not just an intellectual property issue. Companies that use third-party content (web crawl data, publisher APIs, social media data) to train AI systems should audit whether their data licensing arrangements include compensation mechanisms for rights holders — and document those arrangements explicitly.
This is not merely a defensive play. As the Commission develops its enforcement theory in the Google case, companies that can demonstrate fair compensation arrangements for training data will be differentiated from those that cannot. The €2.95 billion September 2025 Google fine for ad tech conduct establishes the financial scale that competition enforcement can reach.
3. Monitor the DMA Scope Review Outcome
The Commission’s consultation on whether to expand the DMA’s core platform service categories to include generative AI closed in May 2026. The outcome of that review — expected in the second half of 2026 — will determine whether AI services face the DMA’s ex-ante obligations (which are more prescriptive than Article 102’s abuse-of-dominance standard) or continue to be handled case-by-case under traditional competition law.
Companies with significant EU AI market presence should monitor the review outcome and assess how a potential DMA expansion would affect their business model. The Commission’s AI Office is also developing its enforcement strategy for general-purpose AI models, with a report expected in Q1 2026 — creating a third regulatory track running parallel to both the DMA review and the ongoing antitrust investigations.
The Bigger Picture: Antitrust as AI’s Real-Time Regulator
The parallel Meta and Google investigations reveal a structural characteristic of the current AI regulatory moment: regulation is racing to catch up with a market that is moving faster than the legislative process. The EU AI Act addresses system-level safety and transparency. The DMA addresses ex-ante obligations on designated platforms. But neither instrument was designed to address the real-time competitive dynamics of AI integration — the question of who controls distribution, training data, and default settings in an AI-enabled market.
Traditional competition law is stepping into that gap. The Article 102 investigations into Meta and Google are using a 70-year-old legal framework to address technology that was commercially non-existent 5 years ago. That approach has limits — case-by-case enforcement is slower than the market and produces bespoke remedies rather than systemic rules — but it is the tool currently available while legislators update the rulebook.
For AI companies operating globally, the lesson is that market power in AI will face competition scrutiny wherever that power is exercised. The EU is moving first and most aggressively. Other major jurisdictions — the United Kingdom’s CMA, the United States DOJ and FTC — are watching the Brussels cases closely and will draw on their findings in their own enforcement decisions.
Frequently Asked Questions
What is the Digital Markets Act and why isn’t it being used in the Meta and Google AI investigations?
The Digital Markets Act (DMA) is the EU’s platform regulation framework for designated “gatekeepers” — large platforms meeting specific thresholds. It imposes ex-ante obligations that companies must comply with regardless of whether specific anticompetitive conduct is proven. However, the DMA’s core platform service categories were defined before generative AI was commercially significant, and AI services like AI chatbot integrations and AI Overviews are not explicitly listed as covered services. The Commission chose to use Article 102 TFEU — traditional competition law — because it is “much more effective” for the specific conduct at issue, while a DMA scope review to include AI services is ongoing.
What would an open-access remedy in the Meta WhatsApp AI case mean for the AI industry?
If the Commission finds that Meta’s API restrictions are anticompetitive and orders open access, it would require Meta to allow third-party AI assistants to integrate with WhatsApp — potentially on the same terms as Meta AI. This would create a major distribution channel that is currently exclusive to Meta AI. Companies with strong AI assistant products would be able to reach WhatsApp’s 2+ billion users through integration rather than competing separately for messaging market share. Open-access remedies of this type have precedent in the EU (interoperability orders in music streaming cases), though implementation timelines are typically 12-24 months.
How does the Google training data investigation affect companies that use web-crawled data to train AI?
The investigation specifically targets Google’s use of publisher content and YouTube data in circumstances where the rights holders were not compensated and not given an option to refuse. The competition law theory — that a dominant company is appropriating third-party inputs without fair terms — applies most directly to companies with dominant market positions. Smaller AI companies training on publicly available web data under standard licensing terms face different risk profiles. However, the case is establishing legal precedents around training data sourcing that are likely to influence regulatory guidance across multiple jurisdictions over the next 2-3 years.
—
Sources & Further Reading
- Regulating AI: Between the EU Digital Markets Act and Competition Law — Hausfeld
- 2026 Antitrust Year in Preview: Big Tech — WSGR
- The Trends That Will Define European Antitrust in 2026 — ProMarket
- Antitrust & Competition Technology Year in Review 2025 — Goodwin
- Key EU Competition Law Developments 2025 and 2026 Predictions — Quinn Emanuel
🔗 Related Intelligence
EU AI Act August 2026: The High-Risk AI Compliance Playbook Before the Enforcement Deadline
EU AI Act August 2026: The Annex III High-Risk Compliance Checklist for Enterprises
EU AI Act August 2026: The High-Risk Compliance Checklist for Non-EU AI Providers
