Home → Chatbot Compliance
EU AI Act Chatbot Compliance: What You Must Do
Article 50 transparency obligations for chatbots, voice assistants, and conversational AI. Deadline: 2 August 2026.
TL;DR: If your chatbot interacts with users, you must clearly disclose it's AI-powered. This is a limited risk obligation under Article 50 — no full compliance documentation needed, just transparency. Penalty for non-compliance: up to €15M or 3% of turnover.
What chatbot operators must do
- Disclose that users are interacting with an AI system (Article 50(1))
- Make the disclosure before or at the start of the interaction
- Use clear, understandable language — not buried in T&Cs
- If your chatbot generates content, mark it as AI-generated (Article 50(2))
- Ensure AI literacy for staff operating the chatbot (Article 4)
Compliant disclosure examples
✅ "You're chatting with an AI assistant. A human agent is available if needed."
✅ "This is an automated AI response. Type 'human' for a real agent."
✅ "Powered by AI. Responses are generated automatically."
❌ "Terms of service apply. See section 14.3 for AI disclosure." (not clear enough)
When a chatbot becomes HIGH-RISK
A standard customer support chatbot is limited risk. But it escalates to high-risk if it also:
- Makes or assists employment/HR decisions (Annex III, category 4)
- Performs credit assessment or insurance pricing (Annex III, category 5)
- Determines eligibility for public benefits (Annex III, category 5)
- Performs biometric identification (Annex III, category 1)
- Recognises emotions in workplace/school settings (prohibited — Art. 5)
EU AI Act chatbot compliance: complete guide
Which chatbots are affected?
Article 50(1) applies to any AI system "designed to directly interact with natural persons." This includes:
- Customer support chatbots (website, app, messaging platforms)
- Sales and lead qualification bots
- Voice assistants and IVR systems
- AI-powered email response systems
- Internal employee helpdesk bots
What about ChatGPT, Copilot, and similar tools?
If you integrate a third-party LLM (ChatGPT, Claude, Gemini) into your product or workflow, you are a deployer. The transparency obligation applies to you — you must disclose to your users that they're interacting with AI, even if the underlying model is from a third party.
Compliance checklist for chatbot operators
| Requirement | Article | Deadline | Effort |
|---|
| Add AI disclosure to chatbot interface | Art. 50(1) | 2 Aug 2026 | 1 hour |
| Review and update Terms of Service | Art. 50(1) | 2 Aug 2026 | 2-4 hours |
| Mark AI-generated content as detectable | Art. 50(2) | 2 Aug 2026 | 1-2 days |
| Train staff on AI literacy | Art. 4 | Already in effect | 1-2 days |
| Document compliance measures | General | Ongoing | Half day |
Penalties
Non-compliance with Article 50 transparency obligations carries fines of up to €15 million or 3% of worldwide annual turnover. For SMEs, fines are proportionally reduced to the lower of the fixed amount or the percentage.
Further resources