EU AI Act Article 50 Final Guidance —
AI Agents Must Disclose Identity and Principal

On July 20, 2026 — just 13 days before Article 50 becomes enforceable — the European Commission released its final 51-page guidance on the AI Act's transparency provisions. For the first time, it provides a definitive answer to one of the industry's most pressing questions: what must an AI agent disclose when it acts on behalf of an enterprise?

The answer is concise and consequential: AI agents must simultaneously disclose their artificial nature and the principal they represent. This is not an optional "powered by AI" badge — it is a comprehensive transparency requirement covering identity, delegation, and accountability.

Compliance deadline: August 2, 2026 — 11 days from today. Maximum penalty: €15 million or 3% of global annual turnover (whichever is higher).

Article 50(1): The Four-Element Test for Direct Interaction

The guidance breaks down the direct-interaction disclosure obligation under Article 50(1) into four elements: an AI system must exist, there must be an intended interaction, a direct exchange, and the counterparty must be a natural person. A single prompt and reply can constitute an "interaction" — and an AI-initiated message seeking a human response qualifies too. Interaction can occur through text, voice, email, physical action, or any channel permitting genuine responsive exchange.

Critical distinction: A back-end tool that sends suggestions to a customer-service employee does not necessarily communicate directly with the customer. Machine-to-machine calls between agents whose output is not intended to reach a person also fall outside this disclosure duty. But the analysis changes when the AI itself sends the message, makes the call, or acts in an environment perceived by a person.

Dual Disclosure Obligation for AI Agents

The guidance goes beyond a bare AI-origin notice for agents. The Commission states that agents capable of making bookings, managing correspondence, negotiating or concluding contracts, or executing purchases should disclose both their artificial nature and the person on whose behalf they act. The stated reason is to make the origin of the interaction, delegated authority, and accountability for resulting actions visible.

This means enterprises deploying AI sales agents, customer service agents, or negotiation agents need their agents to not only say "I am AI" but also "I represent Company X." This dual disclosure obligation has the greatest impact on:

• AI-powered outbound telemarketing systems
• Contract negotiation agents
• Procurement agents acting on behalf of enterprises
• Customer service escalation and complaint handling agents
• Agents sending legal notices or formal communications

Multi-Agent Routing: Disclosure State Must Persist

A key innovation in the guidance addresses multi-agent architectures. When a system includes multiple specialized agents, the disclosure obligation does not stop at the first interface. The Commission explicitly states that where a provider cannot reliably know in advance whether an agent will reach a person, the agent should be designed and instructed to disclose itself whenever human interaction is reasonably likely. This means disclosure state must survive routing decisions rather than depend on the first interface alone.

In practice, enterprises need to maintain five runtime state elements: the agent's artificial identity, the principal it represents, the authority scope for the current action, the human interaction boundary, and evidence that the required disclosure occurred.

Transition Period Misconception: December 2 Does Not Apply to Interaction Disclosure

A critical common misunderstanding: the limited transition period to December 2 applies only to the Article 50(2) machine-readable marking obligation — and only for systems already on the market before August 2. It does NOT apply to the Article 50(1) direct-interaction disclosure obligation. From August 2, all newly placed interactive AI systems must meet the full interaction disclosure requirements. Systems already on the market before August 2 also need to satisfy Article 50(1) immediately — no transition.

Practical Impact on Enterprises

For enterprises deploying AI agents, the final guidance translates into five concrete actions:

1. Map every agent interaction channel: Which agents might reach natural persons? Through what channels (phone, chat, email)?
2. Identify the legal entity and principal: Who is the provider and deployer? Which legal entity does the agent represent?
3. Implement disclosure state propagation across routing: In multi-agent workflows, disclosure information must be passed between agents.
4. Establish disclosure verification mechanisms: Ensure every human interaction has recorded evidence of disclosure.
5. Create non-compliance contingency plans: What happens if an agent fails to disclose correctly?

For enterprises using third-party AI agent platforms, note that the guidance assigns primary compliance responsibility to system deployers. This means even if the agent is developed by a third party, the enterprise deploying it remains responsible for the disclosure obligation.

Outlook

The Commission describes the final guidance as a "first interpretation" that may be revised as implementation experience, enforcement actions, and CJEU rulings develop. But this does not mean enterprises can wait. The August 2 enforcement date for Article 50(1) is fixed, and the four-element test and dual disclosure obligation described in the guidance provide the first clear picture of how enforcement will proceed.

For the AI agent industry, this guidance marks the inflection point from "technically feasible" to "legally compliant." Agents must not only be capable of completing tasks — they must be capable of demonstrating who they are, whom they represent, and who is accountable for their actions.

References

• TECHi: "EU AI Act Makes AI Agents Say Who Sent Them" — https://www.techi.com/eu-ai-act-ai-agents-disclose-who-sent-them/
• Technology.org: "EU AI Act: What Actually Applies on 2 August 2026" — https://www.technology.org/2026/07/17/eu-ai-act-what-actually-applies-on-2-august-2026/
• Jones Walker LLP: "Yes, August 2 Still Matters" — https://www.joneswalker.com/en/insights/blogs/ai-law-blog/yes-august-2-still-matters-the-eu-approved-a-high-risk-ai-delay-but-most-trans.html
• Astraea Counsel: "What US AI Companies Owe by August 2, 2026" — https://astraea.law/insights/eu-ai-act-august-2026-us-companies