The EU AI Act by audience and legal role

If you use AI purely privately, you are not a deployer: the definition expressly excludes use in the course of a personal non-professional activity.

Only deployer obligations are excluded. The regulation continues to apply to providers and professional deployers of the systems you encounter - many of its rules protect you.

When parents use an AI service privately for the family, this is likewise a personal non-professional activity. Children and parents do not become deployers as a result.

AI systems that exploit vulnerabilities of a person or group due to their age, disability or a specific social or economic situation, materially distorting their behaviour, are prohibited where this causes or is reasonably likely to cause significant harm. Age is therefore an express protected characteristic, particularly for children.

Also prohibited are subliminal, purposefully manipulative or deceptive techniques that appreciably impair the ability to make an informed decision and thereby cause or are reasonably likely to cause significant harm.

New and only from 2 December 2026: AI systems that generate or manipulate realistic intimate or sexually explicit depictions of an identifiable person without explicit consent (point (ba)), and systems that generate or manipulate material within the meaning of Article 2(c) and (e) of Directive 2011/93/EU on child sexual abuse (point (bb)), become prohibited.

Providers must design AI systems that interact directly with people so that you are informed you are interacting with AI - unless this is obvious to a reasonably well-informed person from the circumstances.

Anyone professionally deploying deepfakes must disclose artificially generated or manipulated content. Deepfakes are image, audio or video content resembling existing persons, objects, places, entities or events and falsely appearing authentic or truthful. Not every generated medium is a deepfake. The information must be clear at the latest at the first interaction or exposure.

Any natural or legal person with grounds to consider that the regulation has been infringed may submit a complaint to the relevant market surveillance authority.

Using a system under point (ba) or (bb) is only prohibited where the deployer uses it for the purpose of generating or manipulating such material. For point (bb), a 'without right' defence under national law remains unaffected.

A manipulation that does not increase the exposure of intimate parts or alter the nature of sexually explicit activities does not count as manipulation under point (ba).

As soon as you use an AI system professionally under your own authority, you are a deployer - even as an individual. The exception for personal non-professional activity then no longer applies.

If you develop an AI system or have it developed and place it on the market or put it into service under your own name or trademark - for payment or free of charge - you are a provider.

Providers and deployers take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of the systems on their behalf - taking into account knowledge, experience, education, context of use and affected persons. No specific level of AI literacy of any individual has to be guaranteed; no particular course or certificate is prescribed.

As a deployer you must disclose deep fakes. For evidently artistic, creative, satirical or fictional works, the obligation is limited to disclosing the existence of generated or manipulated content in an appropriate manner that does not hamper the display or enjoyment of the work.

As a deployer you only have to disclose AI-generated or manipulated text where it is published to inform the public on matters of public interest. Not every AI-assisted text has to be labelled.

Machine-readable marking of synthetic audio, image, video and text outputs is an obligation of the providers of the generating systems - not of deployers.

If an agency builds a chatbot for clients and puts it into service under its own name, it must, as provider, ensure people are informed about the AI interaction.

No disclosure obligation for text where the AI-generated content has undergone human review or editorial control and a natural or legal person holds editorial responsibility for the publication.

The provider obligation to mark outputs does not apply to the extent the system performs an assistive function for standard editing or does not substantially alter the input data or its semantics.

For generating systems placed on the market before 2 August 2026, providers have until 2 December 2026 to comply with Article 50(2). This does not postpone the other transparency obligations.

A company using an AI system under its own authority is a deployer. Team leads are not a separate legal role; they act on behalf of the deployer.

Whether a system is high-risk does not follow from the sector or company size, but from its intended purpose under Article 6 and the use cases in Annex III.

As deployers you take measures to support the AI literacy of your staff and others acting on your behalf. No specific individual level of literacy or certificate course is prescribed.

Using AI systems to infer emotions of natural persons in the workplace is prohibited. Emotion recognition systems are defined by reference to biometric data - a mood survey in which employees answer themselves is therefore not automatically equivalent.

Anyone lawfully deploying an emotion recognition or biometric categorisation system must inform the persons exposed of its operation.

AI systems intended for recruitment or selection are high-risk - in particular for targeted job advertisements, analysing and filtering applications and evaluating candidates.

Likewise systems for decisions on terms of work, promotion or termination, for allocating tasks based on individual behaviour or personal traits, and for monitoring and evaluating performance and behaviour.

Deployers of such systems use them in accordance with the instructions for use and assign human oversight to persons with the necessary competence, training, authority and support.

Before putting such a system into service at the workplace, employers inform workers' representatives and the affected workers.

Deployers of Annex III high-risk systems that make or assist decisions about people inform those people of the use.

Article 86 applies since 2 August 2026. A person subject to a deployer decision based on the output of an Annex III high-risk system (except point 2) may obtain a clear, meaningful explanation of the AI role and main decision elements: the decision must produce legal or similarly significant effects and be considered by that person to adversely affect their health, safety or fundamental rights. This is not a right to an explanation of every AI output. Separately, Chapter III Sections 1-3 apply to Annex III from 2 December 2027; Article 113 does not categorically postpone Article 86 to that date.

The workplace emotion prohibition does not apply where the system is intended to be put in place or on the market for medical or safety reasons.

An Annex III system is exempt only if it poses no significant risk to health, safety or fundamental rights of natural persons, including not materially influencing decision outcomes, AND meets at least one condition: (a) a narrow procedural task; (b) improving a completed human activity; (c) detecting decision patterns without replacing or influencing the completed human assessment without proper human review; (d) a preparatory assessment task. Systems profiling natural persons always remain high-risk.

The right to explanation does not apply where Union or national law provides exceptions, and only to the extent it is not otherwise provided for under Union law.

An educational institution using an AI system under its own authority is a deployer - this also covers public authorities and bodies.

Education is not high-risk across the board. What matters are the intended purposes listed in Annex III point 3.

The institution as deployer takes measures to support the AI literacy of its staff and others acting on its behalf, such as teachers. No specific individual level or certificate is prescribed.

Using AI systems to infer emotions of natural persons in education institutions is prohibited.

Systems intended to determine access or admission or assign persons to institutions (a), evaluate learning outcomes - including to steer the learning process (b), assess the appropriate level of education (c) or monitor and detect prohibited behaviour during tests (d) are high-risk.

Deployer obligations then apply: use in line with the instructions, human oversight by competent and authorised persons, and informing the persons affected.

Public-law bodies and private entities providing public services must assess the impact on fundamental rights before first deploying relevant Annex III high-risk systems (except point 2). Article 27(1) also covers deployers of point 5(b)/(c) systems. The assessment covers processes, duration/frequency, affected groups, specific risks, oversight and mitigation/complaint mechanisms. Existing assessments may be used in comparable cases; changes require updates. Results must be notified to the market surveillance authority using the template; Article 27(3) includes the Article 46(1) exception.

Learners, especially children, are protected by the prohibition on exploiting age-related vulnerabilities to materially distort behaviour with significant harm.

Deployers inform people of decisions or decision assistance using Annex III high-risk AI (Article 26(11), from 2 December 2027). Separately, Article 86 applies since 2 August 2026: a deployer decision based on such output with legal or similarly significant effects gives a right to explanation where the person considers it to adversely affect their health, safety or fundamental rights; the exceptions in Article 86(2)/(3) remain.

Learners and parents with grounds to consider an infringement may submit a complaint to the market surveillance authority.

The emotion prohibition does not apply where the system is intended for medical or safety reasons.

An Annex III system is exempt only if it poses no significant risk to health, safety or fundamental rights of natural persons, including not materially influencing decision outcomes, AND meets at least one condition: (a) a narrow procedural task; (b) improving a completed human activity; (c) detecting decision patterns without replacing or influencing the completed human assessment without proper human review; (d) a preparatory assessment task. Systems profiling natural persons always remain high-risk.

Anyone who develops an AI system or has it developed and places it on the market or puts it into service under their own name or trademark is a provider. Whether an integrated model comes from a third party does not change this; what matters is your own AI system.

Anyone who provides a general-purpose AI model themselves is subject to Chapter V, which applies since 2 August 2025.

A system is high-risk as a safety component or product under the harmonisation legislation in Annex I requiring third-party conformity assessment - these rules only apply from 2 August 2028.

Article 6(1a) excludes exclusively non-safety aspects of user assistance, performance optimisation/efficiency, automation, usability and quality control from safety components. Under (1b), systems whose failure or malfunction would endanger health and safety are safety components. Under (1c), third-party assessment required solely for other risks unrelated to health and safety does not satisfy (1)(b).

Systems with an intended purpose listed in Annex III are high-risk - these rules only apply from 2 December 2027.

If you consider an Annex III system not high-risk under Article 6(3), you document the assessment before placing it on the market, register under Article 49(2) and provide the documentation to authorities on request.

Article 16(a) requires compliance with Articles 8-15: risk management, data and data governance, technical documentation, record-keeping, information for deployers, human oversight, accuracy, robustness and cybersecurity. The respective conditions apply to the particular high-risk system: Annex III from 2 December 2027, Annex I from 2 August 2028 (Article 113(3)(c)).

Article 16(c)-(i) requires a quality management system, retention of documentation and automatic logs under the provider's control, conformity assessment before placement on the market/putting into service, an EU declaration of conformity, CE marking and registration under the specified articles. Duties apply to Annex III from 2 December 2027 and Annex I from 2 August 2028. Transitional cases remain subject to Article 111(2).

Distributors, importers, deployers or third parties are only considered providers of a high-risk system with the Article 16 obligations in three cases: they put their name or trademark on a high-risk system already on the market (a), make a substantial modification after which it remains high-risk (b), or change the intended purpose of a non-high-risk system so that it becomes high-risk (c). An API integration is not in itself one of these cases.

On a role change, the original provider ceases to be the provider of that specific system and supports the new provider through cooperation, information, technical access and other assistance. This support is expressly excluded if the original provider clearly specified that its system must not be changed into a high-risk system; in that case it also need not hand over documentation.

Systems interacting directly with people must be designed so that people are informed of the AI interaction, unless this is obvious.

Systems generating synthetic audio, image, video or text content mark their outputs in a machine-readable format and detectable as artificially generated or manipulated. For systems placed on the market before 2 August 2026, this obligation applies from 2 December 2026.

Research, testing and development prior to placing on the market or putting into service are excluded - but testing in real-world conditions is not.

AI systems and models specifically developed and put into service for the sole purpose of scientific research and development are excluded.

Systems released under free and open-source licences are excluded - unless placed on the market or put into service as high-risk systems or as systems under Article 5 or 50.

High-risk systems placed on the market before Chapter III applies are only covered if their design changes significantly afterwards; systems intended for use by public authorities must comply by 2 August 2030 at the latest.

Providers of general-purpose AI models placed on the market before 2 August 2025 must comply with their obligations from 2 August 2027.