WhatsApp is still the channel where most small and mid-sized businesses across Latin America, Spain and Italy handle customer service and close sales. It's also where, over the last two years, plenty of businesses bolted on a generative-AI bot without checking too closely what was actually allowed. That changed: since January 15, 2026, a new Meta policy has been in force that redefines what kind of AI can run on the WhatsApp Business Platform. If your company — or the agency that built your bot — hasn't reviewed this yet, you could be breaking the terms without knowing it, with the real risk that WhatsApp simply stops delivering your messages.
What the new policy bans
The rule, which already applied to new accounts starting October 15, 2025, bans using the WhatsApp Business Platform (the API businesses use, not the free consumer app) to offer general-purpose AI assistants. That covers open-ended, ask-me-anything chatbots, integrations that replicate the experience of ChatGPT, Gemini or Perplexity inside the chat, and any system that uses user conversations to train or improve a language model. Meta's logic is straightforward: WhatsApp Business exists so a company can talk to its own customers, not so an AI provider can distribute a generic assistant for free by riding on that infrastructure.
What's still allowed — and worth building
AI automation hasn't disappeared from WhatsApp, it's been scoped down to the business running it. Agents with a specific function tied to the company's own catalog or service are still allowed: customer support about your own products, order tracking, appointment and booking scheduling, transactional notifications (confirmations, reminders, shipping updates), and recommendations drawn from your own catalog. The difference isn't technical — both cases can run on the same underlying language model — it's about scope: a bot that answers from your knowledge base and your products is different from an assistant you can ask anything about any topic.
Why Meta made the change
Meta is under regulatory pressure in Europe over how it handles competition inside WhatsApp, with antitrust investigations ongoing, but the reason the company gave is operational: general-purpose bots were generating a huge volume of messages without producing any revenue for Meta's business-messaging product. The practical result is that, inside WhatsApp, Meta AI is now the only general-purpose conversational assistant, and everything else has to be tied to a concrete task for the business running it.
How to check whether your automation is compliant
If your business already runs AI on WhatsApp, it's worth doing a quick review. Key questions: does the bot answer things outside your catalog or service, like "help me with my homework" or "explain this topic to me"? Are you running a language model with no system prompt or guardrail scoping it to your business? Are you storing or forwarding conversations to train your own model or a third party's? If the answer to any of those is yes, there's a real compliance risk. The fix isn't dropping AI from WhatsApp — it's narrowing it: a clear system prompt, a knowledge base limited to your own catalog and services, and handoff to a human for anything outside that scope.
What this means for your business
For most small businesses, this doesn't change the plan — it clarifies it: automated WhatsApp support is still one of the most cost-effective ways to scale without adding headcount, as long as it's properly scoped. If you're not sure whether your current bot complies with the policy, or want to build one that's compliant from day one, we can audit your current automation or design a new one aligned with the rules in force.