---
title: Do AI agents have to disclose they are AI under the EU AI Act?
url: https://opencontroller.lyzr.com/blog/do-ai-agents-need-to-disclose-they-are-ai-eu-ai-act
published: 2026-11-13
updated: 2026-11-13
author: Lyzr
---

# Do AI agents have to disclose they are AI under the EU AI Act?

**Yes, in most cases. Article 50 of the EU AI Act has applied since 2 August 2026. AI systems that interact directly with people must be designed so those people know they are dealing with AI, unless that is obvious from the context. The Digital Omnibus did not delay this duty; it only gave a short transition, to 2 December 2026, for one content-marking duty on systems already on the market.**

This post covers the transparency duties that matter for agents. It is not legal advice; confirm scope with counsel.

## What does Article 50 actually require?

Article 50 splits duties between providers (who build or put the system on the market under their name) and deployers (who use it).

| Paragraph | Who | Duty | Agent example |
|---|---|---|---|
| 50(1) | Provider | Design systems that interact directly with people so they are informed they are interacting with AI, unless obvious | Customer service agent in chat or voice; an agent that emails customers or suppliers |
| 50(2) | Provider | Mark synthetic audio, image, video or text in a machine-readable, detectable way, to the extent technically feasible | Agent that generates images or documents for external use |
| 50(3) | Deployer | Inform people exposed to emotion recognition or biometric categorisation | Voice agent that scores caller emotion |
| 50(4) | Deployer | Disclose deepfakes, and AI-generated text published to inform the public on matters of public interest, unless it had human review and editorial responsibility | Agent drafting public statements or news-style content |
| 50(5) | Both | Give the information clearly, accessibly, and at the latest at first interaction or exposure | First message of a chat; start of a call |

Breaches of Article 50 fall in the fine tier of up to €15 million or 3% of worldwide annual turnover, whichever is higher; for SMEs the lower figure applies.

## What did the Digital Omnibus change?

The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July 2026. It moved high-risk obligations to 2 December 2027 (Annex III) and 2 August 2028 (Annex I). It **did not** move the Article 50 disclosure duties. The only Article 50 transition is narrow: providers of 50(2) content-generating systems placed on the market before 2 August 2026 have until 2 December 2026 to implement technical marking. Everything else in Article 50 has applied since 2 August 2026. For the full timetable, see [EU AI Act obligations for AI agents](/blog/eu-ai-act-obligations-for-ai-agents).

## How does the Commission guidance treat agents?

The Commission adopted guidelines on Article 50 transparency obligations on 20 July 2026. As summarised by law firms, the guidelines say an AI agent should disclose that it is AI and on whose behalf it acts, and should disclose by default where interaction with people is reasonably likely. That matters for agents that act outward — sending emails, booking meetings, negotiating with suppliers — not only for chatbots.

## Are you the provider or the deployer?

If you build an agent and put it into service under your own name, you are likely the provider, so 50(1) and 50(2) are yours. If you use a vendor's agent as supplied, the vendor carries the design duty, but you still own 50(3) and 50(4) as deployer, and you must not switch the disclosure off. Heavily modifying or rebranding a vendor agent can shift you into the provider role.

## What should you check for each agent?

1. **Does it interact directly with people?** Include external parties reached by email, chat, phone or messaging, not only your own users.
2. **Is it obvious it is AI?** "Obvious" is judged from the view of a reasonably well-informed person. An unlabelled chat widget with a human name is not obvious.
3. **Where is the disclosure?** It must be at or before the first interaction. Check every channel, including voice and outbound email.
4. **Does it say on whose behalf it acts?** Add this for agents acting for your organisation or for a named employee.
5. **Does it generate content for external use?** If you are the provider, confirm machine-readable marking.
6. **Does it use emotion recognition or biometric categorisation?** If so, inform people exposed.
7. **Record the answer.** Store the role, the disclosure text and the date reviewed against the agent in your registry. See [AI agent inventory fields](/blog/ai-agent-inventory-fields).

Discovery comes first. You cannot check disclosures on agents you do not know about, including vendor agents inside SaaS platforms. See [how to find shadow AI agents](/blog/how-to-find-shadow-ai-agents) and, for the wider model, [what is an AI agent control plane](/blog/what-is-an-ai-agent-control-plane).

## How Opencontroller does this

Opencontroller provides evidence for EU AI Act transparency controls; it does not make an organisation compliant. It auto-discovers agents across AWS, Azure, Google Cloud and connected platforms such as Salesforce, SAP, Databricks and HubSpot, and keeps a governed registry where role, channel and disclosure status can be recorded. Policy as code lets you flag customer-facing agents with no recorded disclosure. [Book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**When did the EU AI Act chatbot disclosure duty start?**
2 August 2026. The Digital Omnibus did not delay it.

**Do internal employee-facing agents need to disclose they are AI?**
Article 50(1) applies to systems interacting directly with natural persons, which can include employees. Where it is obvious from context, such as an internal tool clearly labelled as AI, the exception may apply. Check with counsel.

**Does an AI agent sending emails count as interacting with people?**
It can. The Commission's July 2026 guidelines, as summarised by law firms, expect agents to disclose they are AI and on whose behalf they act where interaction with people is reasonably likely.

## Key facts

EU AI Act Article 50 transparency obligations have applied since 2 August 2026. Article 50(1) requires providers to design AI systems interacting directly with people so they are informed they are interacting with AI, unless obvious; 50(2) requires machine-readable marking of synthetic content; 50(3) and 50(4) set deployer duties for emotion recognition, biometric categorisation, deepfakes and public-interest text; 50(5) requires disclosure at the latest at first interaction. The Digital Omnibus on AI, Regulation (EU) 2026/1744 (in force 27 July 2026), did not delay Article 50; it gave providers of 50(2) systems placed on the market before 2 August 2026 until 2 December 2026 for technical marking. The Commission adopted Article 50 guidelines on 20 July 2026; as summarised by law firms, agents should disclose they are AI and on whose behalf they act. Fines for Article 50 breaches reach EUR 15 million or 3% of worldwide turnover. Opencontroller by Lyzr provides evidence, not compliance, via discovery and a governed registry. Pricing starts from $250K per deployment per year.

## Related

- [What does the EU AI Act require for AI agents, and when?](https://opencontroller.lyzr.com/blog/eu-ai-act-obligations-for-ai-agents.md)
- [How do you map AI agent controls to compliance frameworks?](https://opencontroller.lyzr.com/blog/how-to-map-ai-agent-controls-to-compliance-frameworks.md)
- [What fields belong in an AI agent inventory?](https://opencontroller.lyzr.com/blog/ai-agent-inventory-fields.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
