When ChatGPT vanished from WhatsApp in mid-January, 50 million customers misplaced entry in a single day. So did companies that had spent months constructing customer support workflows round third-party AI chatbots on the platform.
Meta’s clarification was simple: the WhatsApp Enterprise API wasn’t designed for standalone AI assistants. The corporate was merely implementing up to date phrases it had introduced in October. The one AI that stayed was Meta’s personal.
It appeared like a routine platform coverage change. Brussels noticed it in another way.
Two months after opening an investigation, the European Fee despatched Meta a Assertion of Objections and threatened emergency motion. The velocity alone is uncommon. However what issues extra is the query regulators are forcing into the open: when a communications platform embeds its personal AI, does it must let rivals in too?
That query doesn’t cease at WhatsApp. Each main UC and collaboration vendor is making related decisions proper now about how open their AI layer will probably be.
Brussels strikes at Silicon Valley velocity
The European Fee doesn’t often work this quick. Antitrust investigations usually drag on for years — usually so lengthy that markets have already moved on by the point enforcement lands.
On 8 February, the Fee despatched Meta a Assertion of Objections and notified it of doable interim measures. This primarily warning it might order Meta to reverse the coverage quickly whereas the broader investigation continues. This got here simply two months after the Fee opened formal proceedings on 4 December 2025.
In keeping with Teresa Ribera, the EU’s competitors chief:
“AI markets are growing at speedy tempo, so we additionally must be swift in our motion.”
Translation: Brussels just isn’t ready years whereas the aggressive panorama will get locked in.
Meta’s defence reveals the true query
Meta isn’t conceding. It says the WhatsApp Enterprise API was constructed for business-to-customer messaging — order confirmations and appointment reminders, for instance — not as a distribution channel for standalone AI chatbots.
“There are various AI choices and other people can use them from app shops, working techniques, gadgets, web sites, and trade partnerships,” a Meta spokesperson mentioned, including that “the Fee’s logic incorrectly assumes the WhatsApp Enterprise API is a key distribution channel for these chatbots,” in line with Reuters’ reporting on the EU’s transfer.
It’s not an unreasonable argument on its face. However it exposes the query each platform vendor will quickly face: the place’s the road between integrating your individual AI and locking out everybody else’s?
That query issues for WhatsApp as a result of — regardless of the “enterprise” label — it capabilities like infrastructure for buyer communication in lots of markets. And it issues much more as a result of WhatsApp just isn’t the one platform making an attempt to turn into the default residence for AI assistants.
This isn’t nearly WhatsApp
Look throughout the communications stack and also you’ll see the identical sample taking part in out in every single place: AI is changing into the interface, and platforms are incentivised to make their AI the default.
Microsoft has Copilot embedded all through Groups. The corporate has additionally signalled help for third-party brokers through Mannequin Context Protocol (MCP), permitting brokers to work with third-party apps and companies in Groups environments. However Copilot stays the privileged expertise, and the licensing mannequin creates sturdy incentives to maintain customers contained in the Microsoft AI ecosystem.
Zoom’s AI Companion is equally baked into conferences, chat, and make contact with centre experiences. Integrations exist, however the default pathway is Zoom’s personal AI — which is precisely the place a platform vendor desires consumer consideration, utilization telemetry, and monetisation to move.
Salesforce, in the meantime, is constructing a complete agent platform with Agentforce because the orchestration layer. Third-party AI doesn’t get equal billing — it operates inside the platform’s framework or in no way.
None of those platforms have confronted the identical scrutiny Meta is now beneath. However the precept the EU is making an attempt to determine — that dominant communications platforms can’t privilege their very own AI by excluding rivals — doesn’t cease at WhatsApp. If interim measures are imposed and the precedent sticks, each vendor embedding proprietary AI right into a dominant platform might want to reply the identical query Meta is going through: how open is open sufficient?
As one authorized evaluation put it within the context of Meta’s WhatsApp integration, platform house owners face a structural trade-off: “An open ecosystem maximises general worth however limits the dominant agency’s capability to seize worth, whereas a closed ecosystem maximises worth seize however will increase innovation prices,” as mentioned on the Kluwer Competitors Regulation Weblog. That trade-off simply grew to become much more costly to get unsuitable.
What to look at, and what to do
Meta will reply to Brussels inside weeks. An oral listening to follows, then a choice on whether or not to impose interim measures. That call will matter for each platform vendor watching to see the place the road will get drawn.
Past this case, the EU’s Digital Markets Act revision is due in Could 2026, and up to date authorized commentary suggests AI is firmly on the agenda.
A Quinn Emanuel round-up of EU competitors developments notes stakeholder help for bringing AI-related companies into the DMA’s scope and flags the rising depth of AI-linked competitors enforcement (see “Key EU Competitors Regulation Developments: 2025 Overview and 2026 Predictions”).
In case your technique will depend on a single platform’s API (whether or not that’s WhatsApp Enterprise, Groups, Zoom, or another communications layer) you’re now uncovered to 2 distinct dangers: the platform can change its phrases, and regulators can power these phrases to vary again. Both consequence can break integrations you’ve already deployed.
The reply isn’t to keep away from embedding AI into your UC stack. The productiveness beneficial properties are actual, and customers anticipate it. However the Meta case strengthens the argument for platform variety.
The EU’s choice on interim measures will arrive in weeks, not years. What follows will decide whether or not UC platforms can deal with AI integration as a product choice or whether or not openness turns into a regulatory requirement. For anybody managing communications infrastructure, that’s not an summary query — it’s a variable in your subsequent platform analysis.
Learn extra:
