For a few months, Meta ran an experiment in what it means to own a doorway. In October 2025 it rewrote the WhatsApp Business terms so that general-purpose AI assistants — the OpenAIs, Anthropics, Perplexities, and Mistrals of the world — would no longer be allowed to operate through the platform’s API. The change took effect on January 15, 2026. From that date forward, the only AI assistant you could summon inside the world’s most-used messaging app was Meta’s own: Meta AI, baked into the interface, impossible to remove, and now without rivals.
On June 10, 2026, the European Commission told Meta to undo it. The order is blunt: reinstate access to the WhatsApp Business Solution for third-party AI assistant providers under the same terms that applied before October 15, 2025 — and do it within five working days. Fail, and the Commission has signaled it will reach for daily penalty payments and fines. In the regulatory architecture Brussels has built around companies of Meta’s size, the theoretical ceiling on such fines is striking: up to 10% of total global annual turnover under the Digital Markets Act, rising to 20% for repeat offenders.
What the numbers actually are
It is worth being precise, because the headline figures get repeated loosely. Meta’s full-year 2025 revenue came in around $201 billion — somewhat higher than the ~$187 billion sometimes cited, though either number makes the same point: 10% of Meta’s turnover is a sum measured in the tens of billions, a deterrent designed to be felt rather than absorbed.
The legal scaffolding deserves the same care. The June 10 action is technically an interim measures decision under EU competition law — the Commission’s preliminary conclusion is that Meta is “likely abusing its dominant position” by refusing rivals access to WhatsApp, an Article 102 abuse-of-dominance theory. But it is inseparable from the Digital Markets Act. WhatsApp is a designated core platform service, Meta is a designated gatekeeper, and the DMA’s entire premise — that a handful of firms control the chokepoints of digital life and must keep them open — is what gives this enforcement its teeth and its 10%-of-turnover ceiling. The “five days” is the deadline attached to the interim order; treat the precise count as a working-week figure rather than a calendar talisman, since interim measures run on working days and can shift with procedure.
One detail captures Meta’s strategy better than any of this. After the initial ban drew scrutiny, Meta revised its policy in March 2026 to technically readmit third-party assistants — but attached a fee. The Commission’s read was that the fee was, in practice, equivalent to the original ban: a toll set high enough that the door is open in name only. That is the move Brussels is now trying to close off. Reopening a gate and then charging an impassable price to walk through it is a tactic gatekeepers know well.
The competition story, and why privacy people should care
On its face this is an antitrust matter. Meta controls the messaging layer — billions of installed conversations, network effects no rival can replicate — and the worry is that it will use that control to win the next market, general-purpose AI assistants, before that market has even settled. Lock competitors out of the channel where people increasingly want to talk to an assistant, and you don’t have to build the best assistant. You just have to own the room.
But the reason this belongs on a privacy blog is that “the room” is your private correspondence. The question the Commission is adjudicating — which AI assistants get to operate inside WhatsApp — is, viewed from your end of the conversation, the question of which companies get to embed themselves in your chats. When Meta blocked rivals, it wasn’t just protecting market share; it was ensuring that the one assistant with a standing presence in your messages was its own. When Brussels orders the door reopened, it is, with the best of competitive intentions, inviting more third parties into that same intimate space.
Neither outcome is automatically good for you. A monopoly assistant is a single point of surveillance and a single set of terms you cannot negotiate. A crowded field of third-party assistants is a wider attack surface, more entities with a foothold near your messages, and more consent screens that nobody reads. The competition remedy solves the lock-in problem and quietly enlarges the privacy one.
Interoperability versus the encrypted envelope
There is a deeper tension the AI fight previews. WhatsApp’s signature feature is end-to-end encryption: in principle, not even Meta can read the contents of your messages. The DMA’s other great demand on Meta — messaging interoperability under Article 7, the requirement to let rival messengers exchange messages with WhatsApp users — has always sat uneasily with that promise. You cannot bolt a third network onto an encrypted system without deciding whose keys, whose servers, and whose threat model govern the seam where they meet. Security researchers have warned for years that mandated interoperability risks weakening the very guarantee that made the platform trustworthy.
The AI-assistant case is a narrower version of the same problem. An assistant operating inside your chat is, by definition, a party to the conversation — it reads what you send it. Opening that channel to many providers is less a breach of encryption than a multiplication of the people standing inside the envelope with you. The right question is not only whether rivals can plug in, but on what privacy terms: what they may retain, what they may train on, whether their access is opt-in and revocable, and whether “interoperable” ever becomes a euphemism for “your messages, now readable by more companies.”
Brussels is right that one firm should not own the doorway to an emerging market by virtue of owning your messages. That is the correct instinct, and the 10% hammer is what makes gatekeepers listen. But interoperability is a competition remedy, and privacy is a separate value that competition law only glances at sideways. The genuinely good outcome is not simply more AI assistants in your private chats. It is a regime where you decide which ones — if any — get a seat, on terms you can see and undo. Forcing the door open is the easy half. Making sure that what walks through it answers to you, and not merely to the next gatekeeper in line, is the part the five-day clock does not address.
Sources: European Commission press corner (IP/26/805); Digital Watch Observatory; HPP analysis; Meta Q4/FY2025 results; European Commission — About the DMA.
