Since 2 August 2026, European law requires in principle that a person knows they are interacting with an artificial intelligence system. The announcement must be clear, accessible and made no later than the first interaction. This transparency replaces neither the identification of the trader nor the information owed on personal data.
Law verified as at 16 August 2026. This page is not individualised legal advice. Article 50 of the AI Act(opens in a new tab) covers systems intended to interact directly with natural persons. How the information is given depends on the channel, the context and whether the interaction with an AI is already obvious — or not.
Yes. Since 2 August 2026, an AI system intended to interact directly with a person must in principle inform that person that they are dealing with an AI, no later than the first interaction. The exception for cases where this is already obvious depends on the context and must be used with caution. This transparency comes in addition to the identification of the trader and, where data is processed, to the GDPR information. Source: Article 50 of the AI Act(opens in a new tab).
No later than the first interaction with the natural person.
No. It depends on what a reasonably well-informed, reasonably observant and circumspect person understands in the actual context.
Yes, in a sales call to a consumer; the commercial nature of the call must also be announced from the outset.
Source : French Consumer Code (code de la consommation), Article L221-16(opens in a new tab)
Every statement on this page links to the text or decision it rests on. The links go straight to the official document.
Editorial author and publisher: Blue Lemon Agent, a brand of LINDBERGH FORMATION. Method: texts and decisions verified on EUR-Lex, Légifrance, the CNIL, the courts and the competent authorities. Legal review: Mohamadou Hamady DIA, juriste, on 18/08/2026. Updated on 10/09/2026.
The detail, rule by rule
Article 50, paragraph 1, of the AI Act(opens in a new tab) concerns AI systems intended to interact directly with natural persons. It requires providers to design and develop them in such a way that people are informed that they are interacting with an AI.
The key wording is this: people must be informed that “they are interacting with an AI system”, unless this is clear to a reasonably well-informed and reasonably observant and circumspect person, taking into account the circumstances and the context.
The text places the design obligation on the provider, but the deployer must make sure that the disclosure mechanism is actually switched on in its own customer journey. A contractual clause does not repair a silent interface.
The exception is not a general exemption for chatbots. It depends on how a reasonably well-informed and reasonably observant person perceives the exchange in the actual context. A robot icon, an explicit label and a plainly automated interface may make the nature of the exchange obvious. A very natural voice, a first name and a script simulating a human adviser produce the opposite effect.
Recommended implementation rule: announce the AI even when the team believes it is obvious. The informational cost is low and the announcement avoids an uncertain assessment after the fact.
For a sales call to a consumer, Article L221-16 of the French Consumer Code (code de la consommation)(opens in a new tab) requires the trader to state, at the beginning of the conversation, its identity, the identity of the person on whose behalf it is calling, and the commercial nature of the call. If the consumer objects, the call must stop immediately and that consumer must not be contacted again.
A compliant opening must therefore answer three distinct questions:
If the exchange involves personal data — telephone number, identity, voice, content of the conversation, customer history — then Articles 13 or 14 of the GDPR(opens in a new tab) also apply. A two-layer notice is generally appropriate: the essentials in the interface or at the beginning of the call, then a link or an option leading to the full notice.
Example of a first layer: “This exchange with the AI assistant of [company] may be recorded and processed for [purpose]. To find out about your rights or to speak to a person, say ‘privacy’ or ‘adviser’.” The wording must match the processing actually carried out; do not announce a recording if there is none, and do not omit it if there is one.
Article 50 of the AI Act(opens in a new tab) has applied since 2 August 2026. Systems placed on the market before that date do not benefit from a general exemption from informing the person they interact with.
Regulation 2026/1744 adds only a targeted transitional measure: providers of systems generating synthetic content placed on the market before 2 August 2026 have until 2 December 2026 to comply with the technical marking obligation laid down in Article 50, paragraph 2, of the AI Act(opens in a new tab). That transitional measure does not postpone the announcement required by paragraph 1 during a direct interaction.
Failure to comply with Article 50 of the AI Act(opens in a new tab) falls under Article 99, paragraph 4: an administrative fine of up to 15 million euros or 3 % of total worldwide annual turnover, subject to the specific calculation rules applying to SMEs and to the proportionality of the penalty.
Source : Official decision (opens in a new tab)
An automated dialler made very short calls to prompt the recipients to call back a premium-rate number. The court upheld a conviction for a misleading commercial practice, in particular because the person on whose behalf the practice was carried out was not clearly identifiable.
Useful takeaway: in an automated journey, identifying the trader cannot be pushed back onto the user.
Limit: this ruling does not decide on Article 50 of the AI Act(opens in a new tab) and does not hold that a customer service must reveal its automated nature. As at the closing date of this handbook, no published and consolidated French court decision has been identified on the precise omission of that announcement.
Article 50 of the AI Act(opens in a new tab) provides for an exception where the AI nature is clear from the context. In practice, an explicit announcement at first contact remains the safest option, especially if the system uses a first name, a natural voice or a human avatar.
No. The information must reach the person no later than the first interaction and must be clear, recognisable and accessible.
Yes, where the system interacts directly with a natural person. The professional capacity of the party on the other end does not set aside Article 50 of the AI Act(opens in a new tab).
Not necessarily. The test is that of a reasonably well-informed and reasonably observant person in the actual context. A voice on its own is a weak indicator.
Yes, in a sales call to a consumer. More broadly, the identity of the controller must also be provided where the GDPR applies.
No, if that means the information is given only after the interaction has actually begun. It must appear in the first screen, the first message or the opening announcement.
The free audit runs your project against the applicable obligations, before it goes live.
