The right to speak to a human when dealing with an AI: what the texts really say

Three families of texts are regularly cited: the GDPR (the EU General Data Protection Regulation, known in France as the RGPD), which governs automated decisions; the European regulation on artificial intelligence (the AI Act), which requires transparency and, for certain systems, human oversight; and consumer law, which governs the relationship with customers. Each answers a different question.

Confusing them leads to two opposite errors: believing that a voice agent is prohibited unless a human is behind it, or believing that no rule applies. Let us take them one by one.

Article 22 of the GDPR gives the person the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning them or similarly significantly affects them. Where this type of decision is permitted, because it is necessary for a contract or based on explicit consent, the controller must at least guarantee the right to obtain human intervention, to express their point of view and to contest the decision.

A voice agent that provides information, takes a message or offers an appointment does not generally take this type of decision. The question arises differently if you entrust it with a choice that significantly affects the person, for example dismissing a request without human review. The right reflex is not to entrust it with that type of decision.

Article 50 of the AI Act requires providers to design AI systems intended to interact directly with people in such a way that those people are informed that they are interacting with an AI, unless this is obvious from the point of view of a reasonably well-informed, observant and circumspect person. The information must be clear and recognisable and be given at the latest at the time of the first interaction. According to the European Commission, this article has applied since 2 August 2026; its guidelines of 20 July 2026 give the example of a voice assistant that announces, at the start of the session, that it runs on AI.

This is a transparency obligation, not a right to a human contact: it allows the caller to know who they are talking to, and therefore to ask for something else if they wish.

Article 14 of the AI Act concerns human oversight, but it targets AI systems classed as high-risk. It provides that natural persons must be able to oversee the system while it is in use, in order to prevent or minimise risks to health, safety or fundamental rights.

This oversight is entrusted to people responsible for supervising the system on the organisation’s side, not to the caller: it therefore does not create, for the caller, a right to be transferred. Whether a use counts as high-risk depends on the cases listed in the annexes to the regulation; if you are in doubt about yours, have its classification checked.

In France, law no. 2025-594 of 30 June 2025 changed the rules on telephone canvassing: since 11 August 2026, telephone canvassing has been prohibited in principle, unless there is an existing contract or the consumer has given prior consent. It concerns outbound prospecting calls, not access to an adviser when the customer is the one calling.

A sector-specific rule does exist, however, for distance contracts for financial services (insurance, loans, savings, for example). Since 19 June 2026, article L222-5-1 of the French Consumer Code (code de la consommation), which comes from ordinance no. 2026-2 of 5 January 2026, provides that when the professional uses online tools, the consumer must be able to address a human being and talk with them, before the contract is concluded and, afterwards, where this is necessary for the consumer’s proper understanding or for the contract’s performance. Outside that case, we did not identify, in the texts consulted for this article, any general obligation to offer a human contact in customer service. Other sectors may have their own rules: if yours is one of them, check them with your legal adviser.

Even without a general obligation, leaving a way out to a person is good protection: announce the nature of the agent from the start, accept the transfer request at any time without making the caller repeat it and, if nobody is available, take a message with a callback at an announced time. Avoid loops in which the agent rephrases without ever transferring.

Alyas, the AI voice agent, follows this approach: it announces at the start of the call that it is an AI and that the call is recorded, answers from the information you give it, says when it does not know and offers to pass the request on, and transfers a caller who asks to your single transfer number. The transcript and summary of each call remain available in your workspace. A free 10-day trial, with no credit card required, lets you check this journey.