Meta’s WhatsApp AI Gatekeeping Faces Antitrust Reckoning

Meta’s concession follows a policy change introduced on 15 January 2026, when the company restricted conversational AI integrations on WhatsApp to its own assistant, Meta AI, while blocking competing bots from using the messaging platform.

Meta argued that the restriction was partly a technical and operational decision. The company said that rapid proliferation of automated chatbots could strain infrastructure designed primarily for human messaging. WhatsApp handles more than 100 billion messages daily, according to Meta’s most recent platform metrics disclosed during its 2025 earnings cycle.

Allowing large numbers of AI agents - each capable of generating multiple automated replies in a single conversation - could significantly increase computational load, message traffic, and abuse risks such as spam or automated phishing campaigns.

However, regulators and AI developers interpreted the move differently. In their view, the restriction effectively turned WhatsApp - one of the world’s largest communication networks - into a closed distribution environment for Meta’s own AI services.

The European Commission’s concern is that this behaviour may constitute “self-preferencing”, a form of platform conduct in which a dominant digital intermediary privileges its own products while limiting competitors’ ability to reach users.

Meta’s response has been to open the platform - temporarily - to third-party AI developers via the WhatsApp Business API, which already powers customer-service chatbots used by banks, airlines, and e-commerce platforms.

Yet the company intends to charge access fees for AI integrations, a point that rivals argue could still make deployment economically unviable at scale.

Why WhatsApp Matters in the AI Distribution Economy

The regulatory stakes are unusually high because WhatsApp is not merely a messaging app; it is one of the world’s most powerful digital distribution channels.

Sources: Meta investor disclosures (2024–2025), Tencent Annual Report, Telegram founder updates (2024), Statista messaging datasets, TRAI digital ecosystem data.

Meta reported in 2024 that WhatsApp had surpassed 2.7 billion monthly active users globally, making it the most widely used messaging platform after China’s WeChat ecosystem. In several large markets - including India, Brazil, Indonesia, and parts of Europe - the app functions as a primary digital interface for communication, commerce, and customer service.

India alone accounts for more than 500 million WhatsApp users, according to industry estimates cited by Statista and the Telecom Regulatory Authority of India’s digital ecosystem assessments.

For AI developers, integration within such a platform dramatically reduces user-acquisition costs. Instead of persuading consumers to download standalone AI apps, conversational assistants embedded directly in messaging interfaces can interact with users in real time - answering questions, generating documents, booking services, or handling customer support requests.

This distribution advantage explains why technology companies are racing to embed AI into messaging ecosystems. Microsoft has integrated AI assistants into Microsoft Teams and Office productivity tools, while Google is deploying conversational AI across search and Android services.

Meta’s own Meta AI assistant - built using the company’s Llama family of large language models - is already integrated across WhatsApp, Instagram, and Facebook, giving it immediate access to billions of users.

For competitors developing independent AI systems - such as those behind ChatGPT and other conversational models - access to messaging ecosystems could become a decisive distribution channel.

Messaging Apps Are Quietly Becoming the “Operating System” of the AI Economy

One reason regulators view access to WhatsApp as strategically significant is that messaging platforms are beginning to function as the primary user interface layer of the digital economy - a role historically played by operating systems and web browsers.

Apart from ~2.7 billion individual users, Meta reported that more than 200 million businesses worldwide use WhatsApp Business tools, including airlines, banks, telecom providers, and e-commerce merchants that rely on automated messaging systems to manage customer interactions.

The geographic concentration of these networks further amplifies their strategic value. Industry data compiled by Statistaand the Telecom Regulatory Authority of India (TRAI) indicates that India alone hosts roughly 500–550 million WhatsApp users, making the country the largest single messaging market in the world. Brazil, Indonesia, and Mexico together account for another 300–400 million users.

Technology analysts increasingly view messaging platforms as a potential “AI operating system.” In a widely cited 2024 research note, venture investor Benedict Evans observed that messaging ecosystems combine three critical distribution assets rarely found together elsewhere in digital infrastructure: persistent user identity, social graphs, and payment or commerce integration. These attributes allow AI assistants embedded within messaging interfaces to bypass traditional distribution barriers such as app downloads, account creation, and search discovery.

The economic implications are substantial. If conversational AI assistants can operate directly within messaging threads - answering questions, generating documents, scheduling services, or handling customer support - they gain access to billions of daily user interactions. That scale could rapidly turn messaging platforms into the dominant distribution layer for AI services, in much the same way that smartphone operating systems shaped the mobile-app economy during the 2010s.

For regulators in Brussels, the concern is straightforward: if a single platform operator controls access to that layer, it may effectively determine which AI services reach global users.

Regulators Test the Limits of Platform Power

European regulators have become increasingly wary of large digital platforms using their ecosystem control to dominate adjacent markets.

The European Union’s Digital Markets Act (DMA) - fully implemented in 2024 - was specifically designed to prevent dominant “gatekeeper” platforms from restricting interoperability or unfairly favouring their own services.

Although the WhatsApp case is formally being pursued under EU competition law rather than DMA enforcement, the underlying regulatory logic is the same: dominant platforms should not leverage control over user access points to determine which AI services reach consumers.

Competition economists at Bruegel, the Brussels-based policy think tank, have warned that messaging platforms may become the “new operating systems” of the AI economy, controlling access to billions of daily digital interactions.

If that occurs, regulators fear that platform owners could replicate the market power once exercised by smartphone operating systems and app stores.

The Commission’s willingness to consider interim measures in this case signals how seriously it views the risk.

Such measures have been used only rarely in EU competition history, most notably in a 2019 investigation involving Broadcom, where regulators intervened before the full antitrust process concluded.

The Browser and App-Store Wars Offer a Historical Precedent

The legal conflict surrounding AI assistants on messaging platforms follows a familiar pattern in the history of digital markets: technological innovation creates a new distribution channel, and competition disputes soon emerge over who controls access to that channel.

In the late 1990s, regulators confronted a similar issue during the landmark antitrust case United States v. Microsoft Corp. (1998), where the U.S. Department of Justice argued that Microsoft had unfairly leveraged its dominance in personal-computer operating systems to promote its own Internet Explorer browser while restricting rival browsers such as Netscape. Courts ultimately concluded that Microsoft’s bundling strategy constituted an abuse of monopoly power.

Two decades later, the conflict re-emerged in a different technological environment. App-store ecosystems controlled by smartphone platforms became the primary gateway for distributing mobile software. The legal dispute between Epic Games and Apple beginning in 2020 focused on whether Apple’s control over the iOS App Store allowed it to impose restrictive payment rules and limit competition from alternative software distribution channels.

European regulators now believe the artificial-intelligence era may produce a similar dynamic. Messaging platforms, productivity suites, and cloud ecosystems could become the next generation of software distribution infrastructure.

Policy researchers at the Bruegel Institute, a Brussels-based economic think tank, have warned that conversational interfaces embedded within large platforms could function as “AI gateways” - digital chokepoints through which billions of users access artificial intelligence services. If those gateways are controlled by a single company, the operator may gain the ability to favour its own AI products or limit rival access.

For regulators drafting the Digital Markets Act, preventing such gatekeeping behaviour was a central objective. The WhatsApp dispute therefore represents not only a specific antitrust investigation but also a test case for whether regulators can prevent dominant digital platforms from controlling the distribution of AI services in the same way earlier platforms controlled web browsers or mobile apps.

Parallel Investigations Across Jurisdictions

The dispute is not confined to Europe. Competition authorities in multiple jurisdictions are investigating similar concerns about AI access to messaging platforms.

In Italy, the national antitrust authority ordered Meta earlier this year to permit rival AI assistants on WhatsApp while conducting its own investigation into potential abuse of market dominance.

In Brazil, the country’s Administrative Council for Economic Defense (CADE) recently succeeded in reinstating a court injunction requiring Meta to allow competing AI chatbots on the platform. The Brazilian case closely mirrors the European investigation, with regulators examining whether Meta’s restrictions unfairly limited emerging AI competitors.

Meta has indicated that its new policy allowing rival AI integrations will apply in Brazil as well as the European Union.

The convergence of investigations across jurisdictions reflects a broader trend: regulators worldwide are increasingly concerned that AI markets could become vertically integrated ecosystems controlled by a handful of large technology platforms.

The Economic Dispute Beneath the Legal Fight

Not all AI developers are satisfied with Meta’s proposed compromise.

One complainant, The Interaction Company of California, developer of the Poke.com AI assistant, argues that Meta’s pricing model for AI integrations could still make access commercially unviable.

Its chief executive, Marvin von Hagen, has told regulators that the company’s revised policy “simply replaces one anti-competitive restriction with another”, replacing an outright ban with access fees that could make large-scale deployment economically impractical.

The dispute therefore reflects not only a legal argument about platform openness but also a commercial struggle over the economics of AI distribution.

Operating large language models is already expensive. According to research by Stanford University’s Institute for Human-Centered Artificial Intelligence (HAI), the cost of running large AI models - including inference infrastructure - can reach several cents per query depending on model complexity and hardware requirements.

Adding platform access fees on top of these infrastructure costs could significantly raise the marginal cost of delivering AI services within messaging environments.

The Hidden Cost Stack Behind AI Bots on Messaging Platforms

Beyond the legal arguments about platform openness lies a more technical dispute over the economics of running AI assistants inside messaging ecosystems.

Large language models - the software systems powering conversational AI - require substantial computing resources to generate responses. According to the 2025 AI Index Report published by Stanford University’s Institute for Human-Centered Artificial Intelligence (HAI), inference costs for advanced AI models can range between $0.01 and $0.03 per thousand tokens for high-performance systems comparable to GPT-4-class models, depending on hardware efficiency and data-centre optimisation.

Sources: Stanford AI Index 2025, SemiAnalysis LLM cost estimates, Meta WhatsApp Business API pricing documentation, AWS and Azure conversational AI hosting estimates.

When such models are deployed inside messaging environments, additional platform costs are layered on top of compute expenses. Access to the WhatsApp Business API, the primary interface through which third-party services interact with WhatsApp users, typically involves conversation-based pricing structures. Meta’s published pricing framework in several markets places these charges in the range of $0.005 to $0.014 per conversation, depending on geography and category of interaction.

The cost dynamics compound quickly at scale. A typical AI-assisted customer support interaction might involve six to ten messages, including two or three AI-generated responses. Infrastructure costs for generating those responses - particularly when using larger models—may already approach $0.02–$0.05 per interaction. When messaging platform fees, hosting expenses, and monitoring infrastructure are added, the total marginal cost of a single AI conversation can climb to $0.03–$0.10.

For a startup attempting to deploy AI services across messaging platforms, these economics become significant. At 10 million monthly conversations, the operational cost could range from $300,000 to $1 million per month, even before accounting for engineering staff, training infrastructure, or marketing expenses.

Competition economists note that such cost structures can quietly shape market outcomes. A 2024 policy paper by the Centre for European Policy Studies (CEPS) observed that platform access fees layered on top of AI inference costs could function as a de facto entry barrier, favouring companies that already operate large data-centre networks or possess significant capital reserves.

In practical terms, the debate over whether Meta can charge rival AI developers for access to WhatsApp is not merely a legal dispute about fairness. It is also a dispute about whether independent AI developers can economically survive inside the infrastructure of dominant messaging platforms.

A Larger Contest Over AI Ecosystems

Behind the legal dispute lies a broader structural shift in the technology industry.

During the smartphone era, companies such as Apple and Google controlled digital distribution through operating systems and app stores. In the AI era, analysts increasingly believe that integration layers - messaging platforms, productivity tools, and cloud ecosystems - may play a similar role.

Whoever controls these layers will shape which AI assistants consumers interact with daily.

Meta’s strategy is clear: embed AI across its existing social-media infrastructure, turning WhatsApp, Instagram, and Facebook into distribution channels for its own AI services.

Rival AI developers, meanwhile, want regulators to ensure that these digital gateways remain open.

The Infrastructure Strain and Strategic Stakes Behind Meta’s Resistance

Meta’s hesitation to open WhatsApp to competing AI assistants is not driven solely by regulatory friction. The company argues that integrating large numbers of AI chatbots into the messaging ecosystem presents both technical infrastructure challenges and strategic platform risks.

Unlike traditional chatbots used for customer service, modern conversational AI systems can generate large volumes of automated responses within seconds. WhatsApp’s infrastructure-originally designed as an end-to-end encrypted messaging service optimised for human communication-handles enormous traffic already. Meta disclosed during its 2025 developer briefings that the platform processes more than 100 billion messages per day, with infrastructure distributed across multiple global data centres.

Sources: Carnegie Mellon Software Engineering Institute chatbot studies, Meta developer documentation, MIT conversational systems research.

AI assistants could significantly alter that traffic pattern. Human messaging typically involves short exchanges with several minutes between replies. AI systems, by contrast, can generate multiple responses within seconds, increasing message frequency and computational load. Engineers studying conversational systems at Carnegie Mellon University’s Software Engineering Institute have noted that AI-driven messaging interactions can produce three to five times the traffic volume of typical human conversations, especially in customer-support scenarios where automated systems provide multiple suggested responses.

Such dynamics introduce operational risks. Messaging platforms must manage spam prevention, rate limiting, and abuse detection systems designed to prevent automated messaging campaigns. Large-scale deployment of AI bots could complicate these controls, potentially increasing exposure to phishing, misinformation, or automated spam networks.

At the same time, Meta faces a strategic calculation. Over the past three years the company has invested heavily in artificial-intelligence infrastructure, including the development of its Llama family of large language models and the deployment of its in-house assistant Meta AI across WhatsApp, Instagram, and Facebook. Meta’s capital expenditure guidance for 2025 projected $30–35 billion in AI-related infrastructure spending, including new data-centre clusters equipped with Nvidia GPUs.

Opening WhatsApp to competing AI assistants therefore carries the risk of diluting Meta’s own AI ecosystem. If rival AI services—potentially powered by competing models - gain traction inside the messaging platform, Meta could lose both user engagement and the behavioural data that helps train and refine its own AI systems.

The company’s current policy shift- allowing rival AI assistants temporary access while charging platform fees—appears designed to balance regulatory pressure with these strategic concerns. It offers regulators evidence of openness while still preserving Meta’s ability to control the economic and technical conditions under which AI services operate inside its messaging infrastructure.

Why This Matters

The WhatsApp case illustrates how artificial intelligence is forcing regulators to rethink traditional competition frameworks.

Historically, antitrust investigations focused on pricing, market share, or mergers. In the AI era, the more important question may be control over access to users.

Messaging platforms process billions of daily interactions and increasingly function as digital operating systems for communication and commerce. If AI services are restricted to assistants owned by platform operators, competitors may struggle to reach users even if their technology is comparable - or superior.

The European Commission’s ongoing investigation will therefore test a critical principle for the AI economy: whether dominant digital platforms must allow competing AI systems to operate within their ecosystems.

The answer could shape the competitive structure of the global AI industry for years to come.