Basics

The EU AI Act at a glance

The EU's AI Regulation (Regulation (EU) 2024/1689) does not regulate AI across the board, but according to the risk of its use. Its AI literacy obligation affects every company that uses AI systems professionally under its own authority (Art. 4, Art. 3(4)), subject to the exceptions in Art. 2.

One obligation applies to virtually every company that uses AI professionally

The AI literacy of your staff under Article 4, whatever the risk class. Our training covers it, in 26 languages, with a certificate per person and one audit report for the whole company.

from €24.69 per person, net of VAT (no subscription)

Train your team How it works

What is the EU AI Act?

The EU AI Act (Regulation (EU) 2024/1689) is the European Union's AI Regulation: it lays down uniform rules for placing on the market, putting into service and using AI systems in the EU. It entered into force on 1 August 2024 and, as an EU Regulation, applies directly in all member states, without the need for national implementing legislation. Its goal is a trustworthy, safe and fundamental-rights-compliant use of AI in the European single market. The Regulation follows a risk-based approach: the higher the risk an AI application poses to people and society, the stricter the requirements.

Importantly, the EU AI Act has extraterritorial reach. Even providers outside the EU fall under it as soon as their AI systems are placed on the market in the EU or their outputs are used there. The obligations also do not take effect all at once, but gradually, a full timeline is available on our deadlines page.

Which risk classes does the EU AI Act define?

The EU AI Act classifies AI according to its risk, commonly in four levels: unacceptable risk (prohibited), high risk, limited risk and minimal risk. The Regulation itself does not use the term "risk class"; it sets out prohibited practices (Art. 5), high-risk AI (Art. 6), transparency obligations (Art. 50) and separate obligations for GPAI models (Chapter V). Each level is explained in the overview of the risk categories with examples. The classification determines the specific obligations:

Unacceptable risk

Prohibited practices, e.g. social scoring or manipulative systems. Banned since 2 February 2025.

High risk

AI in sensitive areas (e.g. recruitment, credit scoring, critical infrastructure). Strict requirements for risk management, data quality and human oversight.

Limited risk

Transparency obligations, users must know, for example, that they are talking to a chatbot or that content is AI-generated.

Minimal risk

All other AI systems to which neither prohibitions nor high-risk or transparency obligations apply. The AI literacy obligation under Art. 4 applies here too.

Risk classes in the EU AI Act: examples, obligations and dates
Risk classExamplesWhat appliesSince / from
Unacceptable risk (Art. 5)Social scoring, manipulative techniques, emotion recognition in the workplace and in education, untargeted scraping of facial imagesprohibitedsince 2 February 2025, further prohibitions from 2 December 2026
High risk (Art. 6, Annexes I and III)Recruitment, creditworthiness assessment, exams in education, critical infrastructure, AI as a safety component, e.g. in medical devicesProviders: risk management, data quality, technical documentation, conformity assessment. Deployers: use in line with the instructions, human oversight, monitoring (Art. 26)from 2 December 2027 (Annex III) and 2 August 2028 (Annex I)
Limited risk (Art. 50)Chatbots, AI-generated or manipulated content (deepfakes)Transparency: disclose the AI, label contentsince 2 August 2026
Minimal riskSpam filters, translation and spelling aidsno obligations under Art. 5, 6 or 50no separate date
All risk classes: AI literacy (Art. 4)every AI system used professionallymeasures for the AI literacy of the people involvedsince 2 February 2025, new wording since 27 July 2026

Even those who use AI only in the minimal or limited risk category have a fundamental obligation that applies regardless of risk class: the AI literacy of employees. All prohibited practices, the eight high-risk areas of Annex III and typical everyday office applications are on the page EU AI Act risk categories.

Regardless of risk class, the AI literacy obligation under Article 4 applies to almost all companies that use AI, already since 2 February 2025. More on the page Art. 4, AI literacy.

Which risk class does your AI system fall into?

Four questions, in this order, settle the classification. One system can trigger several obligations at once, for example as high-risk AI with transparency obligations.

  1. Prohibited? If the use falls under a prohibited practice under Art. 5, it is not allowed.
  2. High risk? If the AI system is a safety component of a product covered by Annex I, or is itself such a product, and the product must undergo a third-party conformity assessment, or if it is used in an area listed in Annex III, such as employment, education, creditworthiness or critical infrastructure, it is usually high-risk AI. Art. 6(3) sets out exceptions for Annex III systems that pose no significant risk.
  3. Transparency obligations? If the system interacts with people or generates text, images, audio or video, the transparency obligations under Art. 50 apply.
  4. Otherwise minimal risk: If none of this applies, there are no obligations under Art. 5, 6 or 50.

In all four cases the AI literacy obligation under Art. 4 applies.

GPAI: general-purpose AI models

A special role is played by GPAI models (General Purpose AI), large, versatile AI models that power modern chatbots and text or image generators. Since 2 August 2025, they are subject to their own obligations, including requirements for technical documentation, transparency towards downstream providers and compliance with copyright law. For particularly capable models with potential systemic risks, additional requirements apply. For models placed on the market before 2 August 2025, providers have until 2 August 2027 (Art. 111(3)). Fines imposed by the Commission under Art. 101 have been possible since 2 August 2026.

Who is affected? Providers, deployers and more

The EU AI Act distinguishes several roles — each with different levels of obligation:

  • Providers: develop an AI system, or have one developed, and place it on the market or put it into service under their own name or trademark (Art. 3(3)).
  • Deployers: use an AI system under their own authority, except for purely personal, non-professional use (Art. 3(4)).
  • Importers & distributors: import or distribute AI systems in the EU market; for high-risk AI systems they must verify conformity (Art. 23, 24).

Crucially, even those who "merely" use AI are covered as deployers (Art. 3(4), Art. 4), particularly regarding the AI literacy of their staff.

What obligations apply and from when?

The EU AI Act takes effect in stages. The key dates at a glance:

  • 1 August 2024: The Regulation enters into force.
  • 2 February 2025: Prohibited practices (Art. 5) and the AI literacy obligation (Art. 4) apply.
  • 2 August 2025: Obligations for GPAI models (for models placed on the market before this date, transition until 2 August 2027), the governance structure and the penalty provisions take effect, excluding the GPAI fines under Art. 101.
  • 27 July 2026: The Digital Omnibus (Regulation (EU) 2026/1744) enters into force, rewords Art. 4 and postpones the high-risk dates.
  • 2 August 2026: The Regulation applies generally, including the transparency obligations under Art. 50 and the GPAI fines under Art. 101.
  • 2 December 2026: New prohibitions under Art. 5 (non-consensual intimate depictions of real persons, child sexual abuse material); end of the transitional period for labelling under Art. 50(2) for systems placed on the market before 2 August 2026.
  • 2 December 2027: Obligations for high-risk AI systems under Annex III (e.g. HR, education, credit scoring).
  • 2 August 2028: Obligations for high-risk AI systems under Annex I, Section A (products covered by EU harmonisation legislation, e.g. medical devices).

All details and exceptions are explained on our deadlines page.

EU AI Act and GDPR: not either-or

The EU AI Act does not replace the General Data Protection Regulation, both apply in parallel. While the GDPR governs the handling of personal data, the EU AI Act addresses the AI system itself and its risks. Where an AI application processes personal data, companies must comply with both frameworks simultaneously. Sound AI literacy helps employees recognise exactly these interfaces in everyday work.

Penalties for infringements

Infringements of the EU AI Act can trigger substantial fines. Art. 99 scales the fines into three tiers:

  • Up to €35 million or 7% of global annual turnover, only for prohibited practices under Article 5.
  • Up to €15 million or 3%, for infringements of certain provider, deployer and transparency obligations (e.g. Art. 16, 26 and 50).
  • Up to €7.5 million or 1%, for false or misleading statements to authorities.

For providers of GPAI models, Art. 101 provides separate fines imposed by the Commission (up to 3 % or €15 million, whichever is higher). For the AI literacy obligation under Art. 4, the EU AI Act itself provides no fine range of its own; sanctions are governed by the national law of the Member States. For small and medium-sized enterprises and start-ups, the lower of the two amounts serves as the cap for each of the three tiers (Art. 99(6)); since the Digital Omnibus, the same applies to small mid-cap enterprises for the tiers covering other obligations and false statements (Art. 99(6a)). More on the fines page.

What requirements does the EU AI Act place on companies?

The requirements depend on role and risk: every company that uses AI professionally must promote the AI literacy of its staff (Art. 4). Providers and deployers of high-risk AI have additional obligations (Art. 16 and 26), plus transparency obligations, for example for chatbots and deepfakes (Art. 50). In practice, this means:

  • Get an overview of where AI is used in your company.
  • Take measures to promote AI literacy under Article 4, tailored to the role and context of use of your employees.
  • Complete an EU AI Act training course and document it verifiably, so you can later explain the measures to the authorities.
  • Keep the EU AI Act deadlines in view, in particular the new dates for high-risk AI systems on 2 December 2027 (Annex III) and 2 August 2028 (Annex I).

Frequently asked questions about the EU AI Act

When does the EU AI Act enter into force?

The EU AI Act entered into force on 1 August 2024; its obligations apply in stages until 2028. The AI literacy obligation under Art. 4 and the first prohibitions under Art. 5 have applied since 2 February 2025, and the Regulation has applied in general since 2 August 2026. The obligations for high-risk AI follow on 2 December 2027 (Annex III) and 2 August 2028 (Annex I). All dates are on the deadlines page.

Which AI systems are not covered by the EU AI Act?

The main exclusions are AI systems used exclusively for military, defence or national security purposes (Art. 2(3)), AI developed specifically for the sole purpose of scientific research and development (Art. 2(6)), and research, testing and development before placing on the market, except testing in real-world conditions (Art. 2(8)). Natural persons who use AI in a purely personal, non-professional activity have no deployer obligations (Art. 2(10)). AI released under free and open-source licences is excluded unless it is placed on the market or put into service as high-risk AI or as an AI system falling under Art. 5 or 50 (Art. 2(12)). Anyone who uses AI professionally, by contrast, is covered.

Does the EU AI Act apply to small businesses too?

Yes. The AI literacy obligation under Article 4 applies to everyone who uses AI professionally, regardless of company size. For fines, however, SMEs bear a lower burden.

Do I need to register my AI systems somewhere?

A registration requirement exists primarily for certain high-risk systems. If you only use AI in the minimal or limited risk category, the AI literacy of your staff is the primary concern. Further answers can be found on our FAQ page.

Train your team now View the fines

Note: this overview is for information only and does not constitute legal advice.