Chatbots have become one of the most common ways for companies to incorporate artificial intelligence into their operations. Customer service, technical support, lead generation, e-commerce, human resources and internal services are some of their most frequent uses.
Since 2 August 2026, however, the use of certain AI systems that interact directly with individuals entails a specific transparency obligation.
Article 50 of the European Union Artificial Intelligence Act (AI Act) requires people to know when they are interacting with an artificial intelligence system.
For many companies, compliance can be relatively straightforward. But it must be implemented correctly and, above all, before or at the beginning of the interaction.
What does Article 50 require?
Article 50(1) of the AI Act provides that providers must ensure that AI systems intended to interact directly with natural persons are designed and developed in such a way that the persons concerned are informed that they are interacting with an AI system.
There is an exception: this information does not need to be provided where it is obvious to a reasonably well-informed, observant and circumspect natural person, taking into account the circumstances and context of use.
The European Commission has clarified in its guidelines on Article 50 that this exception should be interpreted narrowly.
Therefore, from a company’s perspective, it is safer not to rely simply on the assumption that the user “already knows” that they are interacting with artificial intelligence.
Which chatbots are covered?
The obligation concerns AI systems designed to interact directly with individuals.
The European Commission identifies four elements for determining when the obligation applies:
- there must be an AI system within the meaning of the AI Act;
- it must be designed to engage in a genuine two-way exchange with individuals;
- the interaction must be direct, meaning that the AI system itself communicates with the person rather than an employee using AI as an internal tool;
- the interaction must take place with a natural person.
This includes many of the systems currently used by companies as chatbots, virtual assistants, AI agents or conversational avatars.
By contrast, an AI system that operates solely in the background, without communicating directly with an individual, is not subject, for that reason alone, to this specific obligation under Article 50(1).
Not every automated system is necessarily an AI chatbot
It is important to distinguish between a genuine AI system and certain traditional forms of automation.
A form that collects user data, a menu offering predefined responses or a system that merely executes fixed rules does not necessarily constitute an AI system for the purposes of the Regulation.
For this reason, the first step should not be to indiscriminately place an “AI” notice on every digital tool used by the company.
The first step is to identify the systems used by the organisation and determine which of them fall within the Regulation’s definition of an AI system.
This classification should form part of the company’s AI systems inventory.
When must the notice appear?
This is one of the most important aspects.
Article 50(5) establishes that the information must be provided in a clear and distinguishable manner, at the latest at the time of the first interaction or exposure.
It therefore does not appear sufficient to hide this information solely in:
- the privacy policy;
- the terms and conditions;
- the legal notice;
- a secondary page about artificial intelligence.
The user must be able to know that they are interacting with AI when that interaction begins.
In practice, a good solution is to display the notice directly in the chatbot interface, before the first message or alongside the assistant’s initial message.
Examples of notices that can be used
The AI Act does not require the use of any specific wording. What matters is that the information is clear, distinguishable and understandable.
A simple notice may be sufficient:
“You are interacting with an artificial intelligence assistant.”
A slightly more explanatory formula can also be used:
“This service uses an artificial intelligence system to answer your queries.”
For a customer service chatbot:
“I am a virtual assistant powered by artificial intelligence. I can help you with questions about our products and services.”
For a system in which a human may subsequently intervene:
“You are initially speaking with an artificial intelligence assistant. If necessary, your query may be transferred to a member of our team.”
And for an internal assistant used by employees:
“This assistant uses artificial intelligence to help you find information and prepare responses. Verify the information before using it for professional decisions.”
The latter wording also incorporates a recommendation on use, although that additional warning does not strictly derive from the identification obligation under Article 50.
A notice that would probably not be sufficient
Ambiguous wording that does not allow the user to clearly understand the nature of the system should be avoided.
For example:
“Hi, I’m Laura, your digital assistant. How can I help you?”
If “Laura” is actually an AI system, the expression “digital assistant” may not make it sufficiently clear that the user is interacting with artificial intelligence.
The same problem may arise with expressions such as:
“smart assistant”, “automated assistant” or “digital agent”.
Where there is any doubt, it is much simpler to expressly use the words “artificial intelligence” or “AI”.
Is displaying “AI” or an icon enough?
Companies should not rely exclusively on an icon, a robot symbol, stars or the initials “AI” if the context does not allow the user to clearly understand what they mean.
The purpose of the rule is not for the company to be able to demonstrate that it displayed some graphic element associated with AI. The purpose is for the individual to know that they are interacting with an AI system.
For this reason, a short and explicit sentence will generally provide a safer solution.
What if it is obvious that it is AI?
Article 50 provides that information may not be necessary where it is obvious that the person is interacting with an AI system.
However, this exception should be used with caution.
The fact that a company considers it obvious that its assistant is artificial does not necessarily mean that this will be obvious to all users.
This becomes particularly important where the chatbot:
- uses a human name;
- incorporates a realistic photograph or avatar;
- writes in a highly natural manner;
- uses a synthetic voice similar to a human voice;
- presents itself as a member of the team;
- operates through WhatsApp, a messaging application or another channel where users would normally expect to communicate with people.
The more human the system appears, the more important it becomes to clearly identify its artificial nature.
The issue of third-party chatbots
Many companies do not develop their own chatbots. Instead, they contract an external platform and integrate the system into their website, application or customer service channel.
This requires distinguishing between two essential roles under the AI Act: provider and deployer.
Broadly speaking, the provider is the person or entity that develops an AI system or has an AI system developed and places it on the market or puts it into service under its own name or trademark.
The deployer is the person or entity using an AI system under its authority in the course of a professional activity.
Therefore, a company that integrates an AI solution developed by a third party into its website may primarily be acting as a deployer, while the developer of the solution will normally be the provider.
However, the specific configuration matters. Customising, modifying or marketing a system under the company’s own name or trademark may change the legal analysis.
Transparency does not simply mean adding one sentence
For a company, properly complying with Article 50 should form part of a broader AI governance process.
At a minimum, it is advisable to document:
- which chatbot or AI agent is being used;
- who its provider is;
- the purpose for which it is used;
- which individuals interact with it;
- where the transparency notice appears;
- what wording is displayed;
- when the notice is displayed;
- whether human intervention is available;
- how the user can contact a person where necessary;
- what personal data the system processes;
- and who is internally responsible for supervising its use.
This not only allows the company to comply with the obligation in practice, but also to demonstrate subsequently how compliance has been implemented.
Article 50 does not replace the GDPR
Informing an individual that they are interacting with AI does not automatically satisfy data protection obligations.
If personal data is requested or processed during the conversation, the GDPR must also be considered: legal basis, information obligations, processors, international transfers, data retention and, where applicable, automated decision-making.
These are different obligations that may coexist within the same chatbot.
A notice such as “you are interacting with AI” may fulfil a transparency function under the AI Act, but it does not replace the information that may be required under data protection law.
A simple check for any company
If your company uses a chatbot, virtual assistant or AI agent, it is worth asking five questions:
- Is it actually an AI system within the meaning of the AI Act?
- Does it interact directly with individuals?
- Do those individuals clearly know that they are interacting with AI?
- Are they informed, at the latest, at the beginning of the first interaction?
- Can we document and demonstrate how we have complied with this obligation?
If any of these answers is unclear, the system should be reviewed.
A simple obligation, but one that already applies
Chatbot transparency is probably one of the easiest AI Act obligations to implement from a technical perspective.
It is also one of the most visible.
Since 2 August 2026, the transparency obligations under Article 50 have been applicable. Companies already using chatbots, AI agents or virtual assistants should therefore review their interfaces and ensure that users receive clear information from the beginning of the interaction.
There is no need to turn every conversation into a legal notice.
In many cases, a simple sentence — “You are interacting with an artificial intelligence assistant” — may be the starting point.
What matters is that the user knows.
AI Act compliance advice
At our practice, we advise technology companies and organisations using artificial intelligence on compliance with the AI Act, including AI system inventories and classification, transparency obligations, reviews of chatbots and AI agents, internal policies and compliance documentation.
The objective is not merely to identify which provisions of the Regulation apply, but to translate those obligations into concrete measures that can be implemented and documented within the company.