Almost certainly yes, and this is the single most common misreading of Article 50 among technical teams.
The intuition is understandable: OpenAI (or Anthropic, or Mistral) trained the model, so surely the transparency obligations are theirs. But the EU AI Act draws its line at the system, not the model. The Commission's draft guidelines are explicit that Article 50 applies to AI systems built on general-purpose models, and whoever places such a system on the market under their own name or brand is its provider. Your chatbot, with your name on it, on your site, is your system. The model licence underneath doesn't move the disclosure duty upstream.
So the Article 50(1) duty of informing people they're interacting with AI, at or before first interaction, in the interface itself, is yours to implement. Your model vendor can't do it for you: they never see your UI.
One adjacent duty is worth knowing about if your product doesn't just chat but generates content users take away (text, images, audio, video): the machine-readable marking duty under Article 50(2). There you may lean on what the upstream model provider embeds (watermarks, provenance metadata), but the draft guidelines say you rely on it "without prejudice to" your own responsibility to demonstrate compliance, meaning you must verify the marking actually survives your pipeline.
The compliance cost for the chat part is trivial: a first-message notice and a persistent badge. The cost of getting the role analysis wrong is a disclosure gap that any regulator, journalist, or competitor can verify from the outside in thirty seconds.
Not sure how this applies to your specific setup? The free 2-minute check tells you which of the four Article 50 duty areas likely apply to your company. No email required.
Take the free check Generate disclosure text