WhatsApp AI Chatbot Rules in 2026: Meta's Policy and the EU AI Act
Two rule changes landed in 2026 for anyone running AI on WhatsApp. Meta now forbids general-purpose AI assistants on the WhatsApp Business Platform, and the EU AI Act requires chatbots to tell people they are talking to a machine. Neither stops a business using AI for its own customers — but both change how you design it.
Regulatory content. This is a plain-language summary as of September 2026, not legal advice. Check the current text of Meta's terms and the AI Act before you rely on it.
Meta's general-purpose AI ban (from 15 January 2026)
Meta updated its WhatsApp Business Solution terms so that providers of AI models and assistants cannot use the platform to distribute a general-purpose AI chatbot, where the AI is the main service. Open-ended "ask me anything" bots offered through WhatsApp fall into this category.
What remains allowed is AI that serves your own business:
- Answering questions about your products, orders, bookings and policies.
- Qualifying leads and scheduling appointments.
- Summarising conversations and drafting replies for agents.
- Routing and triage before a human takes over.
The practical test: is the AI incidental to a service your business provides, or is the AI itself what you are offering? Keep the agent scoped to your domain and refuse off-topic requests.
EU AI Act transparency duty (from 2 August 2026)
Article 50 of the AI Act requires that people are informed when they interact with an AI system, unless that is obvious. For a WhatsApp assistant, it is rarely obvious — a message thread looks the same whether a person or a model wrote it. Disclose at the start of every AI-handled conversation.
Designing a compliant WhatsApp AI agent
- Scope it. Ground answers in your own knowledge base and decline unrelated questions.
- Disclose it. First message says it is an automated assistant.
- Offer a human. A clear keyword or button that hands over to the inbox.
- Log it. Keep transcripts and handover events for audit.
- Mind GDPR. Process personal data under a lawful basis, keep it in-region, and cover your AI provider in your DPA and sub-processor list.
How Arino One fits
Arino One's embedded AI runs on your dedicated instance, grounded in your own content, with human handover into the unified inbox and full transcripts. It is built for business AI — the kind Meta still permits.
FAQ
Has Meta banned AI chatbots on WhatsApp?
Meta's WhatsApp Business Solution terms prohibit general-purpose AI assistants — where the AI itself is the product being offered — from 15 January 2026. AI used to support a business's own customer service, sales or operations remains allowed.
What does the EU AI Act require of chatbots?
Article 50 requires that people interacting with an AI system are informed they are dealing with AI, unless it is obvious from the context. These transparency obligations apply from 2 August 2026 under the Act as adopted.
How should a WhatsApp AI agent disclose that it is AI?
State it clearly at the start of the conversation, for example: 'I'm the automated assistant for [brand]. You can ask for a person at any time.' Keep a visible route to a human.
Is this legal advice?
No. This guide summarises public rules as of September 2026. Regulatory timelines can change, so confirm your specific use case with your legal adviser.
Apply this to your own deployment
This guide describes decisions we make on live instances. Tell us your channels, systems and region and we will map it to an architecture outline, a provisioning plan and an indicative commercial model — usually within one business day.