Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act — C(2026) 5054 final, Annex

(30) Four elements must be present for a system to fall within the scope of Article 50(1) AI Act: (i) the system must be an AI system, (ii) intended to interact, (iii) directly, (iv) with natural persons.

i. An AI system: The system must fulfil the elements of the AI system definition (18), which excludes simple non-AI automated response mechanisms (e.g., traditional out-of-office emails, rule-based quick message answers).

ii. Intended to interact: The AI system must be intended to interact with natural persons. Interaction entails the capability of a bidirectional exchange of information or actions between natural persons and the AI system with a genuine conversational or responsive character. It can occur in various forms (e.g. auditory, visual and physical). Natural persons should be able to provide input (e.g., written text or other forms of content, voice, or physical actions) to the AI system (e.g. via technical communication means including online interfaces, phone or e-mail, or directly), while the AI system should be capable of responding with contextual output (including any type of content or actions) and vice-versa. The interaction may cover any type of output or format that can be perceived and understood by humans. It does not have to be initiated by a human; it may also cover AI content or actions eliciting a human response. The interaction may be a one-time exchange (e.g. single prompt followed with a single reply) or take place over a certain period time (multi- turn). Systems that only passively collect data and are not capable of engaging in an exchange with natural persons (e.g. automated facial-recognition access controls) or just collect one-time feedback (e.g. spam filters), are not considered to interact with natural persons.

iii. Direct interaction: The interaction between the AI system and the natural persons must be direct. This typically involves real-time or near real-time interaction (including through capabilities of the system to write and send messages to natural persons or otherwise interact with the physical or virtual environment that can be perceived by natural persons). Direct interaction excludes indirect or mediated human interaction where a natural person is exposed to AI outputs without directly interacting with the system (for example, if customer service representatives use AI assistance tools to help them communicate with natural persons or if the AI output is not made available to the person by the AI system itself, but by another person disseminating the content). However, the mere possibility for humans to intervene or review the AI system’s outputs should not be used to circumvent the application

(18) See Article 3(1) AI Act. For further guidance on the concept of AI system, see Commission Guidelines on the definition of an AI system, 29.7.2025, C(2025) 5053 final.

of the disclosure requirement under Article 50(1) AI Act. Cases of products and services blending AI-generated responses with human-curated content fall within the scope of Article 50(1) AI Act and require disclosure for the specific outputs in the interaction that are AI-generated, unless those AI outputs have been properly reviewed and sent by humans as the main interlocutors with the natural persons.

iv. With natural persons: Article 50(1) AI Act requires that the AI system interacts with natural persons. Those persons may be professional deployers, other users (also using the system for purely personal activities) or other persons using the system on their behalf. AI systems that operate in a closed physical environment (such as part of an industrial machinery set-up), backend machine-to-machine calls between AI systems (including AI agents) whose outputs are not intended to reach natural person, or virtual environments without any direct contact with natural persons are excluded from the scope of the obligation.

(31) AI agents are covered by Article 50(1) AI Act if they are capable of interacting with the persons instructing them or with other natural persons in the execution of the tasks (e.g. making bookings, managing correspondence, negotiating or concluding contracts, executing purchases, etc). AI agents must be designed and developed in such a way that they disclose both their artificial nature and the person on whose behalf they are acting, considering the need for transparency of the origin and the delegation of authority and accountability for the consequences of their actions. This includes cases where AI agents are capable of interacting in complex multi-agent architectures with other AI agents that directly interact with the natural persons in the execution of the task. Nevertheless, it may not always be possible for the provider of the AI agent to identify in advance individual instances of direct interaction of an AI agent with other natural persons once the AI agent is deployed in different contexts. Where the provider cannot reliably determine whether the AI agent will directly interact with a natural person before placing the AI agent on the market or putting it into service, the agent should be designed at the architecture level, and instructed, to disclose itself as such in every situation where it is reasonably likely that the agent may interact with a natural person (including where a natural person may represent or act for a legal entity). Furthermore, AI agents should also disclose themselves, to the persons instructing them at key steps (e.g. at the point of authorisation, reporting, validation etc., including when the deployed AI agent receives, processes, or relies upon outputs generated by other AI systems rather than directly by a natural person) and at every new interaction. Examples of AI systems intended to interact directly with natural persons falling within scope of Article 50(1) AI Act:

- AI-enabled voice assistants, chatbots/conversational agents in various contexts (e.g. public service, customer support, complaints management, e-commerce, finance, healthcare, education etc.),

- AI hotline for incidents, fraud or other reporting, (humanoid) robots/cobots, AI companions; robotic companion pets; AI avatars (e.g. in virtual reality environments),

- AI bots on social networks and media, coding agents and other AI agents capable of directly interacting with natural persons (including in complex multi-agent architectures). 3.2.1Examples of AI systems not directly interacting with natural persons falling outside the scope of Article 50(1) AI Act:

- AI-enabled traditional industrial robots that operate in a closed industrial setting and not intended to interact with humans (e.g. electronic assembly robots),

- AI-enabled recommender systems, spam filters, automated translation or transcription tools,

- AI-enabled search and retrieval of information without generating or modifying content,

- AI-enabled text and code auto-completion techniques, authentication/biometric recognition systems, backend decision-support systems where the user only sees the AI output without possibility for direct interaction with the system;

- AI-enabled predictive maintenance and optimisation systems in factories and industrial applications, AI-assisted product visualisation (e.g., virtual try-on, room staging, product rendering), AI-enabled single-turn route-planning tools, transcription of customer calls processed for behavioural analysis etc.