Approving the Giant, Litigating the Conduct: The Hindsight of EU Merger Control

On Meta/WhatsApp, the AI access dispute, and a decade of optimistic clearances now corrected through Article 102

I. A second charge sheet, twelve years late

Earlier this week, the European Commission sent Meta a fresh Statement of Objections in the WhatsApp/AI access investigation matter. The document is unusual in two respects.

First, it does not address the underlying conduct — the October 2025 amendment to the WhatsApp Business Solution Terms, which barred third-party general-purpose AI assistants from the platform from 15 January 2026 onwards. It addresses, instead, the remedy Meta had offered in March 2026: paid access for one year. The Commission’s preliminary view is that a pay-to-play arrangement reproduces, in another form, the very exclusion the original objections identified.

Second, and more interestingly, the Commission reaffirms its intention to impose interim measures under Article 8(1) of Regulation 1/2003, ordering Meta to restore the conditions in force before 15 October 2025, while the Article 102 TFEU investigation continues. The Italian competition authority had already imposed equivalent national measures in December 2025. The picture, taken as a whole, is one of an enforcer mobilising the entire toolbox of behavioural intervention to keep open a digital interface whose closure was made possible — indeed, made structural — by a merger clearance the same Commission delivered in 2014.

A reader who follows the file will recognise the institutional choreography: a dominant platform leverages a position the Commission itself confirmed years earlier; the Commission, having watched the position consolidate, now intervenes (too late?) through the abuse-of-dominance pillar to engineer access where structural separation was never imposed. The investigation is, as a matter of doctrine, perfectly justified. As a matter of merger policy, it is an admission.

II. The original sin: Case M.7217

The acquisition of WhatsApp by Facebook was cleared unconditionally in October 2014 (Case M.7217). The Commission’s reasoning, in its essentials, was that Facebook’s social network and WhatsApp’s messaging application addressed distinct customer needs, that several alternative messengers were available in the EEA, and that user matching between the two services was, on the notifying party’s representations, technically infeasible.

Three years later, in May 2017, the Commission fined Facebook €110 million for having provided incorrect or misleading information on precisely that point — the technical possibility of automated user matching — both in the notification and, repeated, in its replies to a request for information (Case M.8228). Two infringements, two separate fines, the first ever imposed under Article 14 of the EU Merger Regulation since its 2004 entry into force. The decision was, in its way, a record.

What the decision did not do is what would have mattered most. The Commission did not revoke its 2014 clearance. It explained, in Case M.8228, that its assessment of the transaction had considered “what if” scenarios that included automated matching, and that the false representations had therefore not materially altered the substantive outcome. From an institutional standpoint, this is not a defence; it is an admission. If the Commission had already considered the worst case and cleared the transaction anyway, then either the worst case was not, in fact, the worst case — which the post-2016 evolution of WhatsApp would suggest — or the substantive standard the Commission applied was insufficiently demanding.

The current AI access investigation is, in substance, the resolution of that ambiguity. WhatsApp has become, by the Commission’s own preliminary findings, an important entry point for general-purpose AI assistants to reach European consumers. That entry point exists because, in 2014, an already dominant social-network operator was permitted to acquire its most credible challenger in interpersonal communications, on assurances later proven false. The Article 102 case the Commission is now building is a behavioural correction of a structural decision — a less efficient instrument applied late, after the harm to potential entrants has already crystallised.

III. A pattern in agriculture: Cases M.7932 and M.8084

The Facebook/WhatsApp clearance is not an isolated misjudgment. The same architectural error — a major incumbent absorbs a credible challenger on a market already dominated by a small number of players, and the merger is cleared on the basis of remedy packages and innovation rhetoric — seems to be a familiar presence also in two of the Commission’s most consequential agricultural decisions of the last decade.

In March 2017, the Commission cleared the Dow/DuPont merger (Case M.7932) subject only to the divestiture of the bulk of DuPont’s global crop protection R&D organisation. The combined entity, restructured along the lines that produced Corteva in 2019, joined a market already organised around Syngenta, Bayer, BASF and a handful of others. The official justification was that the divestiture remedy preserved both price competition for existing pesticides and incentives for innovation in safer and more effective products. Whether innovation in fact requires consolidation — and, if so, why structural concentration should be the preferred vehicle for it rather than the contestability that concentration removes — was not seriously addressed in the public reasoning. It was assumed. Worst, the Commission failed to see that large corporations were already shying away from the unpopular business of selling pesticides, made responsible for various harms to the environment, from bees to crops themselves. The focus and the future were on seeds, the essential ingredient of the agribusiness industry.

A year later, in March 2018, the Commission cleared Bayer’s acquisition of Monsanto (Case M.8084), again subject to a remedy package: divestitures in seeds, pesticides and digital agriculture in excess of €6 billion, the bulk of which was eventually purchased by another giant, BASF (Case M.8851). The transaction produced, by the Commissioner for Competition’s own characterisation at the time, the most important global player in seeds and pesticides. The Commission’s substantive analysis again relied on the proposition that the structural concentration of the industry would be offset by remedy-driven competition and innovation. The empirical foundations of that proposition, in either Dow/DuPont or Bayer/Monsanto, were never publicly developed.

The pattern is consistent. A market already exhibiting high concentration is permitted to consolidate further; the challenger — DuPont’s R&D platform, Monsanto’s seed and trait portfolio, WhatsApp itself in 2014 — is absorbed by an incumbent; remedies are calibrated to address the most easily measurable horizontal overlaps; and innovation rhetoric does the rest of the work. The optimism is not necessarily wrong in any individual case. It is wrong as a methodology, because it transforms the burden of proof: the Commission, instead of requiring that consolidation be demonstrated to enhance competition, accepts, relaxed, that consolidation will not, on balance, harm it. That is not the standard the Merger Regulation contemplates.

IV. The counterexample: Case M.10188

Against this background, the Illumina/GRAIL saga (Case M.10188) acquires its full meaning. In April 2021, on a referral from six national competition authorities — none of which had jurisdiction over the transaction under their own national merger control rules — the Commission accepted to review Illumina’s acquisition of GRAIL on the basis of an extended interpretation of Article 22 EUMR. It went on to prohibit the transaction in September 2022, to order its unwinding, and to impose a €432 million gun-jumping fine.

On 3 September 2024, the Court of Justice annulled that entire architecture (Joined Cases C-611/22 P and C-625/22 P, Illumina and GRAIL v Commission). Following the Opinion of Advocate General Emiliou of 21 March 2024, the Court held that Article 22 EUMR does not authorise the Commission to accept a referral from a Member State that itself lacks jurisdiction under its own national rules. The Commission had acted ultra vires. The substantive prohibition decision, the unwinding order and the gun-jumping fine all collapsed for the mere lack of a jurisdictional foundation, something casting a big doubt on how professional the European Commission truly is.

The economics of the case are a matter of some debate; the jurisdiction is no longer. What deserves attention here is not the legal defeat itself, but the contrast it creates with the agricultural and digital decisions discussed above. In Illumina/GRAIL, the Commission strained the limits of its competence to reach a transaction that, by its own admission, did not meet the EU or national thresholds, which basically means that it was of low importance, at best. In Facebook/WhatsApp, Dow/DuPont and Bayer/Monsanto, transactions that fell squarely within its competence and were made among giants or giants and Mavericks were cleared on optimistic assumptions that subsequent events have, with varying degrees of severity, showed to be inacurrate.

The Commission, in other words, has spent a decade applying the Merger Regulation in inverse proportion to the underlying competitive risk: assertive where the case was thin and the jurisdiction contested, indulgent where the case was strong and the jurisdiction unambiguous.

V. The diagnosis: a competence in search of a doctrine

Of all the powers vested in the Commission by the Treaties and by secondary law, merger control is the one that comes closest to genuine industrial policy. Antitrust enforcement under Articles 101 and 102 TFEU operates ex post on conduct; State aid review constrains Member States rather than firms; the Digital Markets Act and the Digital Services Act regulate by category and by designation. Merger control, alone, is the moment at which the Commission exercises a quasi-political judgment about the structure of European markets — about which firms may grow, by which means, and at what cost to whom.

The inconsistency identified above is therefore not a series of unrelated misjudgments. It is a doctrinal gap. The Commission has not articulated, in any sustained form, what it understands by “innovation” in the merger context, nor why innovation should be assumed to follow from consolidation rather than from the contestability that consolidation removes. It has not explained how it reconciles the precautionary stance it adopts in some files with the deferential stance it adopts in others. It has not addressed why, in the digital sector, the structural error of 2014 is now to be corrected through behavioural remedies under Article 102 — a tool that the Commission itself has acknowledged, in successive policy documents, as an inferior substitute for ex ante intervention.

The result is the precise situation the WhatsApp/AI investigation now exemplifies. The Commission, having allowed a dominant gatekeeper to absorb its principal challenger, must now spend years of enforcement resources to police that gatekeeper’s conduct toward a generation of new entrants — entrants whose viability depends on access to a platform that need not have been so dominant in the first place. The behavioural remedy under Article 102, even if ultimately successful, cannot reproduce the market structure that an unconditional clearance in 2014 foreclosed. It can only mitigate, asymmetrically and at considerable institutional cost, the consequences of that foreclosure.

VI. A possible conclusion

The new mandate at DG COMP, with a still fresh commissioner and a newly appointed Director general inherits, among other files, the operational consequences of the Meta/WhatsApp interim measures, the post-Illumina recalibration of below-threshold review, and the broader question of how the Commission proposes to align its merger doctrine with its evolving digital and industrial policy ambitions. The intellectual challenge is not, primarily, one of enforcement priorities. It is one of methodology.

A merger policy that treats optimism as a substitute for evidence will continue to produce decisions that the next decade’s enforcers must undo. A merger policy that confuses jurisdictional assertiveness with substantive rigour will continue to lose at the Court of Justice on the cases that matter least, while losing on the merits — silently, and in the markets themselves — on the cases that matter most. The Commission’s competence in mergers is the closest it comes to shaping European industry. It deserves a doctrine of comparable seriousness.