Welcome to the Competition Stories led by Alice Setari and Mario Siragusa at Cleary Gottlieb Steen & Hamilton. This issue presents recent developments in EU competition law enforcement in digital markets. It examines the growing phenomenon of parallel investigations by the Commission and national competition authorities – illustrated by the WhatsApp and Amazon Buy Box cases – and the challenges of enforcement allocation within the European Competition Network. The analysis also explores the evolving treatment of tying abuses through the lens of the SAP and Microsoft Teams cases, highlighting the unresolved tension between traditional abuse categories and the realities of digital market conduct. Finally, it surveys further enforcement developments, including the Google AdTech fine, the Google content harvesting investigation and the doctrinal implications of the Google Android Auto ruling for the essential facility doctrine.
***
1. Parallel investigations on the rise: enforcement allocation in digital markets
EU competition law enforcement rests on a multilevel architecture: the Commission and national competition authorities share responsibility for applying Articles 101 and 102 TFEU. This decentralised system, introduced with Regulation 1/2003, has significantly strengthened enforcement. But it also raises an increasingly pressing question: how do you prevent fragmentation when the same conduct is under investigation at both the EU and national level simultaneously?
Nowhere is this question more acute than in digital markets. Large platforms operate across the EU, and national authorities have shown a growing appetite for rapid intervention, including through interim measures. The recent investigations into Meta’s conduct concerning artificial intelligence services on WhatsApp bring this issue squarely back to the forefront.
The WhatsApp investigation
On December 4, 2025, the Commission opened a formal antitrust investigation into whether Meta’s new policy on artificial intelligence (AI) providers’ access to WhatsApp breaches Article 102 TFEU (Case AT.41034).
WhatsApp is a messaging service for exchanging text messages, voice notes, photos, videos and documents, as well as making voice and video calls. But it is more than a personal communications tool. Through the WhatsApp Business Solution, businesses can interact directly with customers, and several AI providers have seized on this channel to offer conversational AI services within the app, giving users access to tools that answer questions, generate content or provide automated support.
In October 2025, Meta drew a line. It announced a new policy prohibiting AI providers from using the WhatsApp Business Solution where AI constitutes the primary service offered (WhatsApp Business Solution Terms). Businesses could still deploy AI tools in ancillary roles –automated customer support, for instance– but standalone AI chatbot services offered by third parties through WhatsApp would no longer be permitted.
The Commission has expressed concerns that the new policy could block third-party AI providers from reaching their customers through WhatsApp in the EEA – while Meta’s own AI service, ‘Meta AI’, would remain freely accessible to users on the platform.
Here is where things get interesting. The Commission’s press release made clear that the formal investigation covers the EEA except for Italy – a deliberate carve-out designed to “avoid an overlap with the Italian Competition Authority’s ongoing proceedings for the possible imposition of interim measures concerning Meta’s conduct”.
The reason? The Italian Competition Authority (ICA) had already moved. On 22 July 2025, the ICA opened an investigation into Meta for a possible abuse of dominant position concerning the company’s integration of Meta AI into WhatsApp, where it was given greater prominence than competing services.
Following the announcement of the new WhatsApp Business Solution Terms, the ICA broadened the scope of its proceedings – examining the very same conduct that the Commission would shortly take up. The ICA found that the changes to the contractual terms could limit production, restrict market access and hinder technical development in the AI chatbot services market, to the detriment of consumers, potentially amounting to an infringement of Article 102 TFEU. It did not stop there. On 24 December 2025, the ICA imposed interim measures on Meta, ordering the company to immediately suspend the WhatsApp Business Solution Terms and preserve access to the WhatsApp platform for Meta AI’s competitors.
Given the substantial overlap in conduct, undertakings and market context, one might have expected a different outcome, namely, centralisation of the case at the EU level.
Saverio Valentino, a commissioner at the ICA, has said as much. He acknowledged that the Commission retains the ability to extend its investigation to Italy, and, should it do so, the Italian authority would close its own case and be “very happy to do so”. Those words are telling. They hint at an emerging awareness within the European Competition Network that more coordinated solutions are needed in cases of clear cross-border relevance, not just to enhance enforcement effectiveness, but also to ensure coherence, legal certainty, and a reduction in administrative burdens for undertakings caught in the crossfire of multiple proceedings.
Previous case law from the Court of Justice
The EU Courts have most recently addressed the permissibility of parallel proceedings between the Commission and NCAs in two key lines of case law.
The first is the Slovak Telekom saga – and it set the stage.
On 8 April 2009, the Commission opened proceedings against Slovak Telekom for abusing its dominant position in the Slovak broadband services market (high-speed internet). The very next day, the Slovak competition authority fined the same company €17.5 million for abusing its dominant position in the markets for retail telecommunications and wholesale interconnection (telephone services and low-speed internet).
Slovak Telekom fought back. It challenged the national authority’s decision in the national court, arguing that the initiation of proceedings at the EU level should have barred the national authority from adopting a decision on the same conduct. Specifically, it claimed the approach violated (i) Article 11(6) of Regulation 1/2003, which aims to prevent conflicts of competence between the Commission and NCAs, and (ii) Article 50 of the Charter of Fundamental Rights of the European Union, enshrining the ne bis in idem principle.
On 25 February 2021, the Court of Justice delivered its answer (Case C-857/19), and clarified that Article 11(6) only strips national competence where the Commission investigates the same undertaking for the same conduct, in the same product and geographic market, and over the same period. Since the Commission and the Slovak authority were investigating infringements in different markets –broadband services on one side, retail telecommunications and wholesale interconnection on the other– the Slovak authority retained its competence.
The second relevant precedent –the Amazon Buy Box case– pushed the analysis further.
In April 2019, the ICA opened an investigation into whether Amazon had abused its dominant position by reserving a set of exclusive benefits –essential for gaining visibility and increasing sales on its Italian marketplace (amazon.it)– to sellers that used Amazon’s logistics service. In November 2020, the Commission launched its own investigation into the same conduct, covering the entire EEA except Italy, since the ICA was already on the case.
Amazon challenged this approach, arguing that the exclusion of Italy exposed it to parallel proceedings. Both the General Court (Case T-19/21) and the Court of Justice (C-815/21 P) rejected the appeal. The Court clarified that Article 11(6) of Regulation No. 1/2003 only applies where the Commission and a national authority are pursuing the same undertaking for the same allegedly anticompetitive conduct in the same product and geographical market over the same period. If the Commission has not initiated proceedings in respect of a given territory, the undertaking cannot invoke Article 11(6) as a shield.
The referring Italian Administrative Court (TAR Lazio) drove the point home. It emphasised that (i) the two sets of proceedings –national and EU– were independent of one another, noting that, since the proceedings initiated by the Commission did not involve Italy, Amazon could not have invoked the protection afforded by the aforementioned provisions against the parallel proceedings, (ii) there is no right to a “one-stop-shop” investigation at EU level, and (iii) the possibility of different outcomes is simply “natural” in a system that allows both the Commission and national competition authorities to enforce antitrust rules in relation to the same practice across different territories (Order of 29 January 2024, No. 399).
Taken together, these strands of case law do more than delineate the limits of Article 11(6) and the ne bis in idem principle. They also reveal a clear institutional preference underpinning the EU competition enforcement system: decentralisation is prioritised, even where this entails duplication of proceedings and a heavier compliance burden for undertakings. In this light, parallel enforcement is not an unintended side effect to be minimised, but a structural feature of the system. The Courts appear willing to accept the risk of overlapping investigations, procedural fragmentation and potentially divergent outcomes as the price to pay for a model that maximises enforcement reach, even if this comes at the expense of legal certainty and procedural economy for the undertakings concerned.
The Google AdTech investigation: a better approach?
A comparable situation arose in the Google AdTech investigation, but this time, the procedural outcome was markedly different.
On 20 October 2020, the ICA opened proceedings against Google concerning alleged exclusionary practices in the Italian display advertising market. The focus was on Google’s handling of user identifiers for advertising targeting. The ICA’s theory: Google had engaged in discriminatory practices, refusing to provide competitors with the keys needed to decrypt Google users’ IDs and limiting third-party tracking pixels, while allowing its own internal services to rely on those very same tools. The authority also flagged the strategic importance of data collected across Google’s ecosystem –Android, Chrome, Google Maps, Google ID– which could be leveraged to entrench its position in advertising intermediation markets.
But the Italian investigation never reached a final decision. On 12 October 2021, the ICA closed the case pursuant to Article 11(6) of Regulation 1/2003, after the Commission decided to open proceedings covering the same conduct (Case AT.40670).
The difference is telling. Rather than carving Italy out of its geographic scope, the Commission chose to centralise enforcement at the EU level, thereby avoiding parallel proceedings entirely and ensuring a single, coherent assessment.
This outcome sits squarely within the logic of the Joint Statement of the Council and the Commission on the functioning of the network of competition authorities. Case allocation within the ECN is designed to reflect where the effects of an infringement are felt and where enforcement can be most effective, taking into account “the ability of the authority to gather evidence, to bring the infringement to an end and to apply sanctions effectively” (§ 15). Crucially, the Statement also stresses that “cases will be dealt with by a single competition authority as often as possible”, with a single NCA typically best placed only where one Member State is substantially affected (§ 16). By contrast, the Commission is “particularly well placed to deal with a case if more than three Member States are substantially affected by an agreement or practice, if it is closely linked to other Community provisions which may be exclusively or more effectively applied by the Commission, if Community interest requires the adoption of a Commission decision to develop Community competition policy particularly when a new competition issue arises or to ensure effective enforcement” (§ 19).
Digital markets, however, rarely respect national boundaries. The practices of large platforms are rolled out –and produce effects– across the EEA almost by design. In that context, purely national effects are the exception rather than the rule, and the logic of the ECN framework points decisively towards EU-level intervention. Centralised enforcement by the Commission may therefore not be merely a matter of efficiency, but the natural institutional response to the inherently cross-border architecture of digital conduct.
The increase in the phenomenon of parallel investigations in digital markets
Parallel investigations are becoming increasingly frequent and may raise concerns regarding consistency, duplication and legal uncertainty, particularly in digital markets. These markets are inherently pan-European, and fragmented enforcement risks producing inconsistent outcomes, duplicative procedures and heightened legal uncertainty for companies operating across multiple Member States.
At the same time, however, parallel enforcement is not without its functional advantages. It may foster a degree of regulatory experimentation and enable authorities to respond more swiftly and flexibly to emerging competitive concerns, especially in fast-moving sectors. This creates an inherent tension between, on the one hand, the need for coherent and predictable enforcement and, on the other, the benefits of a decentralised and reactive system.
What is clear is that this tension is likely to intensify. Parallel investigations are increasingly amplified by the growing overlap between competition law and new regulatory instruments.
Germany offers a clear illustration. Since January 2021, Section 19a(1) of the German Act against Restraints of Competition (ARC) empowers the Bundeskartellamt to impose obligations on undertakings with “paramount cross-market significance”, without requiring a full effects-based analysis in the traditional Article 102 TFEU sense. These rules run in parallel with, and overlap with, both (i) general competition law and (ii) the Digital Markets Act (DMA), as the Federal Court of Justice recently confirmed. Major digital platforms –Google, Amazon, Apple, Meta– have already been designated under this regime and may therefore face parallel investigations in Germany under multiple legal bases simultaneously.
A similar overlap exists between antitrust enforcement and the DMA. Since the DMA’s entry into force, the same gatekeeper conduct may simultaneously trigger antitrust and regulatory enforcement by the Commission. This introduces a double layer of complexity, reflecting the complementary but potentially overlapping objectives of ex ante and ex post regulation.
First, as highlighted by Thibault Schrepel and Godefroy de Boiscuillé in “Questioning the Digital Markets Act’s Legality”, the Commission itself may investigate –and ultimately fine– the same conduct under both regimes. Recent cases illustrate this risk of cumulative enforcement. Consider the €1.84 billion fine imposed on Apple in March 2024 for anti-steering practices affecting app developers, such as Spotify and Deezer, followed by a subsequent fine in April 2025 under Article 5(4) DMA for the same anti-steering practices; or the €2.42 billion fine in the Google Shopping case, as well as the more recent €2.95 billion fine in the AdTech case, which clearly overlap with Article 6(5) DMA on self-preferencing.
Second, and most relevant here, parallelism may also arise across institutional levels, with NCAs applying antitrust rules while the Commission enforces the DMA on the same underlying conduct. The Italian Google/Hoda case is a case in point. The ICA examined whether Google had hindered the development of competing data-driven services by restricting effective data portability under Article 20 GDPR. The proceedings closed with commitments, including the introduction of an early adoption programme for a new portability solution –the Data Portability API– enabling, for certain data scopes, direct service-to-service data transfers for third-party operators authorised by end users, covering both user-provided data and data generated through activity on Google’s services. The twist? Google had already been developing this solution to comply with Article 6(9) DMA and rolled out the full portability solution in March 2024, in line with the DMA deadline. The Italian investigation had, in effect, anticipated the application of Article 6(9) DMA –at least for certain data scopes– further illustrating the risk of parallel investigations targeting the same conduct.
The need for stronger allocation mechanisms
Against this backdrop, the allocation of cases within the European Competition Network is no longer a procedural footnote but rather a central issue.
Clearer allocation mechanisms between the Commission and national authorities would reduce duplication, make better use of enforcement resources and strengthen legal certainty for market participants. Stronger coordination would also help defuse potential tensions with the ne bis in idem principle, particularly where parallel proceedings risk leading to multiple sanctions for closely related conduct.
Cases like Google AdTech point the way. Where conduct is inherently cross-border and produces uniform effects across the internal market, early centralisation at the EU level delivers a single, coherent assessment, avoiding fragmentation and reducing the administrative burden on undertakings.
None of this undermines the role of national authorities. Their proximity to markets remains essential, particularly for detecting infringements and addressing conduct with primarily national effects. But in cases with a clear European dimension, especially in digital markets, a more decisive allocation in favour of the Commission is not just desirable, it is necessary.
2. Tying abuses in digital markets: the limits of a traditional abuse category
The WhatsApp investigation discussed above underscores the Commission’s sustained focus on abuse of dominance in digital markets. A look at the other cases pursued during the same period only reinforces that conclusion.
Two recent cases are particularly revealing when it comes to understanding how Article 102 TFEU enforcement is evolving in response to digital market dynamics. Both involve tying/bundling in the cloud software sector and rest –at least formally– on similar theories of harm. Yet the differences between them are striking, and they expose the range of approaches to leveraging conduct in the digital economy.
The SAP case: a traditional aftermarket tie
On one side sits a traditional tying case. On 25 September 2025, the Commission opened an investigation into SAP –the German-based multinational behind widely used business management software– focusing on the aftermarket for maintenance and support services related to its on-premises enterprise resource planning software (ERP software). The Commission provisionally considers SAP to hold a dominant position in that market.
The conduct under scrutiny is wide-ranging, and the pattern is familiar. The Commission is concerned that SAP requires customers to source all maintenance and support services for their on-premises ERP software exclusively from itself, under uniform pricing conditions across all products. It is also investigated whether SAP prevented customers from terminating maintenance and support contracts for unused software licences, extended the initial licence term during which termination is barred, and imposed reinstatement and back-maintenance fees on customers who re-subscribed after a period without SAP support.
This is, in essence, a textbook tying case where the primary product meets the aftermarket. The theory of harm follows a relatively traditional tying framework: the products are distinct (albeit complementary), there is well-defined independent demand for each, and there is no functional interdependence. The Commission is assessing whether SAP is dominant in the primary product market, and whether the bundling forecloses competition in the secondary market by shutting out third-party providers of maintenance and support services for on-premises ERP software.
The Microsoft Teams bundling case: beyond traditional tying
On the other side sits a far more layered case. The Commission accepted commitments in proceedings it had opened against Microsoft concerning the bundling of Microsoft Teams with its broader suite of productivity applications, without finding an infringement.
This case is noteworthy for several reasons. It marked the first antitrust proceeding against Microsoft in more than ten years. The last had concluded in 2013 and concerned non-compliance with commitments previously imposed by the Commission. The reference point, in other words, belongs to a very different era: a time when Microsoft was the primary target of enforcement action, commentators were calling for its break-up, and the most prominent digital case of the period concerned Microsoft’s tying of its operating system with various software applications.
What followed was a prolonged period of détente. Commentators described a transformation of Microsoft “from bad boy to the most Brussels-friendly of tech giants”, a characterisation that becomes all the more striking when set against the multi-billion-euro fines accumulated by other major tech companies through years of sustained confrontation with EU regulators. That détente appears to have survived the present proceedings largely intact: the case closed with the acceptance of Microsoft’s commitments and no fines imposed.
At the centre of the case is Microsoft Teams –Microsoft’s well-known communications and collaboration platform– which has been bundled by default in Office 365 and Microsoft 365, Microsoft’s productivity suites for business customers, since its launch.
The matter landed on the Commission’s desk in 2020, following complaints from Alfaview, a German company, and Slack, a US company. Formal proceedings were opened in 2023. During the investigation, Microsoft modified its commercial policy, offering versions of its suites without Teams at lower price points. The Commission considered these measures insufficient and, in June 2024, issued a Statement of Objections. It expressed concerns regarding the possibility that Microsoft holds a dominant position in the worldwide market for productivity applications for professional use, and that, since at least April 2019, Microsoft has been bundling Teams with its market-leading productivity applications in breach of Article 102 TFEU. While such integration may generate efficiencies and product improvements, the Commission considered whether Teams could have gained an undue competitive advantage in terms of distribution, further reinforced by interoperability limitations between Microsoft’s core productivity applications and competing communications services.
In response, Microsoft offered the following commitments, which the Commission ultimately accepted:
- To offer customers in the EEA (later extended worldwide on a unilateral basis) versions of Office 365 and Microsoft 365 without Teams, at an appreciably lower price than the equivalent bundled suites;
- To give EEA current customers recurrent opportunities to switch to Teams-free suites, with those suites deployable in datacentres worldwide;
- To allow competitors of Teams and certain third parties to benefit from effective interoperability with specific Microsoft products and services;
- To allow EEA customers to extract their Teams messaging data for use in competing solutions.
Two cases, two approaches
The contrast between these two cases lays bare a broader tension in the Commission’s enforcement of Article 102 TFEU in digital markets. SAP involves a textbook aftermarket tie. The Microsoft Teams case, by contrast, appears to belong to a rather different species – a more complex category of conduct whose theory of harm does not map straightforwardly onto traditional frameworks.
In the Teams case, the bundling was achieved through a degree of technical integration between the services. That integration partially transformed the primary product – the secondary service became, in effect, one of its functional components. And it is precisely this kind of dynamic, common in technology markets, that blurs the traditional boundaries between tying, self-preferencing and, to a lesser extent, refusal to deal.
Both characterisations carry a degree of force. There is considerable merit, as the Commission argued in its Statement of Objections, in framing the conduct as an indirect obligation to purchase Teams alongside the Office suite in which Microsoft holds dominance. But there is also a plausible case for treating it as self-preferencing: Microsoft, by facilitating automatic and seamless access to Teams through its other applications, arguably leveraged its dominant position in the productivity market to favour its own communications service. The same conduct, then, lends itself to different theoretical foundations, with potentially divergent implications.
This ambiguity is not unique to Microsoft Teams. It runs through much of the Commission’s digital abuse enforcement, a direct consequence of the fluid and varied nature of leveraging behaviour in digital markets. Time and again, the Commission has faced situations in which the same conduct could plausibly be characterised in more than one way. In some cases –Google Shopping, Apple App Store (Music streaming)– tying dynamics were not emphasised. In others –Google Android, Facebook Marketplace– they were central to the analysis.
A fully settled organising principle has yet to emerge. When Google Shopping was decided, a number of commentators suggested that the pro-competitive and pro-innovation merits of product integration could serve as the distinguishing criterion separating self-preferencing from tying. That criterion has not endured, and the Commission’s selection of the applicable theory of harm remains difficult to predict in cases that fall close to the boundary between multiple categories.
A comparison between the Microsoft Teams and WhatsApp investigations makes the point vivid. The WhatsApp case could, in principle, have been framed as a tying case: WhatsApp as the dominant primary service, Meta AI as the secondary service bundled with it, and the integration foreclosing competition in the AI assistant market. In both cases, the integration of a new service into an existing platform created meaningful synergies and there was arguably good reason for the link to be maintained while being opened to competitors. Yet only one was analysed through the lens of self-preferencing; the other was treated as a tying case. The basis for that distinction is far from self-evident.
The theoretical uncertainty spills over into remedies. Most of Microsoft’s commitments –offering unbundled suites at lower prices, allowing customers to renegotiate long-term contracts– are precisely what one would expect in a classic tying case. But the interoperability measures, requiring Microsoft to open its productivity applications to competing communications services, are remedies more typically associated with self-preferencing theories. The commitments, in other words, straddle both frameworks – reflecting the unresolved nature of the underlying characterisation.
The practical implications for legal certainty are significant. An undertaking whose conduct could plausibly fall under more than one category of abuse has no reliable way of knowing in advance which legal standard will apply, let alone which remedial framework it may face.
Rather than drawing rigid lines between abuse categories, the Commission has opted for a different path: a uniform framework built on general criteria such as competition on the merits. The problem? Those criteria remain insufficiently defined. The 2024 draft Guidelines on Exclusionary Abuses embody this approach, proposing a unified abuse standard grounded in general principles rather than predefined categories of conduct. The public consultation on those guidelines closed well over a year and a half ago, yet the final text remains outstanding. The Commission has signalled that adoption may come in the first half of 2026, though even that timeline is far from certain.
Until that reform materialises, the three cases examined above will continue to coexist: the WhatsApp investigation, pursued as a self-preferencing case; the Microsoft Teams proceedings, treated as something analogous to –but not reducible to– a tying abuse; and the SAP investigation, built on a conventional tying theory. Together, they offer a vivid illustration of the unresolved tension between the flexibility that enforcement demands and the predictability that the rule of law requires.
3. Further enforcement updates
Beyond the cases discussed above, the Commission opened another Article 102 TFEU investigation against Google (AT.40983) on 9 December 2025, alleging that Google may have infringed Article 102 TFEU by harvesting content from web publishers and YouTube uploads to train its AI models.
Separately, the long-running Google Adtech investigation reached its conclusion on 5 September 2025, with the Commission imposing a fine of EUR 2.95 billion. At the heart of the case: self-preferencing in the online advertising market. The Commission found that Google had engaged in practices it considered exploitative in online advertising intermediation to channel business towards its own Google Ad Exchange platform, giving Ad Exchange preferential treatment in publisher-run auctions and compelling advertisers using Google’s ad buying tools to rely exclusively on Ad Exchange.
But it was the Google Android Auto case that generated the most doctrinal controversy. On 29 October 2025, the Italian Council of State delivered its judgment following the Court of Justice’s preliminary ruling of 25 February 2025 in Case C‑233/23. The Court of Justice drew a critical distinction: where a dominant undertaking has voluntarily opened its platform to third parties –rather than developing it exclusively for its own use– the strict indispensability requirement established in Bronner does not apply. In such circumstances, a refusal to grant access to a competitor is capable of constituting an abuse of a dominant position under Article 102 TFEU under certain conditions without the need to demonstrate that access to the platform is indispensable for competing on the downstream market.
The Italian Council of State followed suit. It classified Android Auto as an open platform, thereby dispensing with the indispensability requirement altogether, while nonetheless ordering a recalculation of the fine originally imposed by the Italian Competition Authority.
The implications are far-reaching. While the Court of Justice underscored the importance of preserving investment and innovation incentives –as well as the dominant undertaking’s freedom of contract– as factors that may justify restraint in certain circumstances, the combined effect of these rulings may be read as an expansion of the conditions under which access obligations may be imposed. The threshold for compulsory access is, in practical terms, lowered and this widens the distance between the EU framework and other major antitrust systems, at a moment when calls for deregulation and stronger protection of innovation incentives are gaining particular force.
The reaction from commentators has been swift and largely critical, with many flagging the risk of chilling incentives to innovate. If dominant firms can be compelled to share infrastructure that is neither indispensable nor developed for third-party use in the traditional sense, the boundaries of the refusal-to-deal doctrine become far less predictable, and the space for unilateral business decisions correspondingly narrower.
Ultimately, the evolving case law reflects an attempt to reconcile two competing objectives: preserving competitive opportunities in digital ecosystems while maintaining incentives for firms to invest and innovate. Whether the balance struck by the Courts will prove adequate to that task remains to be seen.
Mario Siragusa, Alice Setari, Francesco Trombetta, Alberto Galasso