Guides · Article 50 · EU AI Act
Does embedding an LLM chat need an AI disclosure?
Dropping a large-language-model chat widget onto a website — an embedded assistant, a docs Q&A box, a sales concierge — is now a few lines of code. The transparency question is less simple. Two paragraphs of the EU AI Act can be relevant: Article 50(1) where the chat interacts directly with people, and Article 50(2) where it generates synthetic text. This guide explains both from the regulation text. It is informational only, not legal advice, and does not name any particular product as in or out of scope.
The interaction limb — Article 50(1)
Article 50(1) provides that providers must ensure AI systems intended to interact directly with natural persons are designed so those people are informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant and circumspect person given the context (Article 50(1), Regulation (EU) 2024/1689). An embedded LLM chat that visitors type to and receive answers from is the kind of directly interactive system this paragraph is written around.
Note the paragraph frames the duty on providers. Whether embedding a third-party model into your own site makes you a provider for that integration — as opposed to a deployer of someone else's system — is a fact-specific role question that turns on how the system is built, branded, and placed on the market. If your setup presents a conversational agent to the public, 50(1) may be in view; confirm your role with qualified counsel rather than assume it.
The synthetic-text limb — Article 50(2)
Because an LLM writes its replies, the content-marking paragraph can also come into view. Article 50(2) asks providers of AI systems generating synthetic audio, image, video or text to ensure the outputs are marked in a machine-readable format and detectable as artificially generated or manipulated, with solutions that are effective, interoperable, robust and reliable as far as technically feasible. It carves out systems performing an assistive function for standard editing or that do not substantially alter the input. Whether a chat integration's generated text falls inside 50(2), and whether a carve-out applies, are fact-specific.
What integrators tend to map
- Whether the chat interacts directly with people (points toward the 50(1) analysis).
- Whether it generates synthetic text, and if so who is the provider of that generation (points toward the 50(2) analysis).
- Whether you are provider-side, deployer-side, or both for the integration.
- Timing and accessibility of any disclosure under Article 50(5).
- Counsel's review of the role and scope mapping before relying on it.
Common questions
If I embed a third-party LLM, am I the provider?
Not automatically. Article 50(1) and 50(2) frame their duties on providers, but whether integrating a third-party model makes you a provider for that setup — versus a deployer — is a fact-specific role question. Confirm with qualified counsel; our free scope check gives an informational starting point.
See what may apply to your business
Answer seven quick questions for an automated, informational indication of which Article 50 obligations appear likely to apply — free, and not legal advice.