EU AI Act Art. 50: Does Your Chatbot Have to Disclose It's AI?

Since 2 August 2026, Article 50 of Regulation (EU) 2024/1689 (the AI Act) has been in force — the so-called transparency obligations. For most small and medium businesses running a chatbot or voice assistant, one thing changes concretely: the person on the other end must be able to tell they're talking to an AI.

What this means in practice: your chatbot has to disclose it's AI at the start of a conversation — not buried in the terms and conditions, but right in the conversation itself. The same applies to voice assistants: the disclosure has to be spoken aloud, not just written somewhere.

AI-generated images, videos or audio content carry a similar duty: they must be marked as artificially generated in a machine-readable way. Deepfakes require an even clearer, human-perceptible label.

What doesn't apply yet: the stricter obligations for high-risk AI systems (Annex III) were pushed back to December 2027 by the Digital Omnibus (EU) 2026/1744. For the vast majority of chatbot and assistant use cases in SMBs, that wouldn't have applied anyway — but the Art. 50 disclosure duty does, and it applies now.

In our own AI assistant, this disclosure is built in and cannot be turned off — not an optional extra, but the default.