Zum Hauptinhalt springen
Law · EU

EU AI Act: Regulation on Artificial Intelligence

Updated on 6 min Reviewed by: Cenedril Editorial
EUEU AI Act

An HR department tests a tool that pre-sorts job applications. The tool comes from a US provider, the contract is signed quickly. What nobody checked during procurement: systems that take part in hiring decisions are high-risk systems under the AI Act, and the organisation that uses them has obligations of its own. The AI Act regulates AI according to the risk its use creates and assigns obligations to every role in the supply chain.

Regulation (EU) 2024/1689 on artificial intelligence (AI Act) is the first comprehensive law on AI. It applies directly in all member states and covers providers and deployers regardless of where they are established, as soon as an AI system is placed on the market or used in the EU or its output is used in the EU. In July 2026, Regulation (EU) 2026/1744 (Digital Omnibus on AI) postponed deadlines, added two prohibitions and eased obligations for small and medium-sized enterprises.

The risk classes

Prohibited practices (Art. 5). Certain uses are banned, for example subliminal manipulation, exploiting the vulnerabilities of vulnerable groups, social scoring, untargeted scraping of facial images from the internet and emotion recognition in the workplace and in education. With the Digital Omnibus, two prohibitions are added from 2 December 2026: systems that generate non-consensual intimate material or child sexual abuse material.

High-risk systems (Art. 6). High-risk are AI systems that are a safety component of a product under harmonisation legislation listed in Annex I, and systems in the areas listed in Annex III: biometric identification, critical infrastructure, education, employment and workers management, access to essential services (such as creditworthiness, insurance), law enforcement, migration, and administration of justice and democratic processes. An Annex III system is exceptionally not high-risk if it does not pose a significant risk (Art. 6(3)); the provider documents this assessment.

Transparency obligations (Art. 50). Anyone interacting with an AI system is informed of it. Synthetic content is marked in a machine-readable way, deepfakes are disclosed.

General-purpose AI models (Chapter V). Providers of such models keep technical documentation, a copyright policy and a summary of the training data. For models with systemic risk, evaluations, incident reporting and cybersecurity are added.

All other AI systems are not subject to specific obligations under the regulation, apart from AI literacy under Art. 4.

Obligations by role

RoleCore obligations for high-risk systems
ProviderRisk management system (Art. 9), data governance (Art. 10), technical documentation (Art. 11), record-keeping (Art. 12), instructions for use (Art. 13), human oversight (Art. 14), accuracy, robustness, cybersecurity (Art. 15), quality management system (Art. 17), conformity assessment, declaration of conformity, CE marking, registration (Art. 43, 47 to 49), post-market monitoring (Art. 72), reporting of serious incidents (Art. 73)
DeployerUse according to the instructions, human oversight by competent persons, input data, monitoring of operation, keeping the logs, informing workers (Art. 26); fundamental rights impact assessment for public bodies and certain private deployers (Art. 27); explanation of decisions on request (Art. 86)
Importer and distributorChecking conformity assessment, documentation and marking before making the system available (Art. 23, 24)

Deadlines

DateWhat applies
1 August 2024Entry into force of the regulation
2 February 2025Prohibited practices (Art. 5) and AI literacy (Art. 4)
2 August 2025Obligations for general-purpose AI models, governance, penalties
27 July 2026Entry into force of the Digital Omnibus (Regulation (EU) 2026/1744)
2 August 2026Application of the remaining provisions, such as the transparency obligations under Art. 50
2 December 2026Two new prohibitions (Art. 5(1) points (ba) and (bb))
2 December 2027Requirements for high-risk systems under Annex III (Chapter III Sections 1 to 3)
2 August 2028Requirements for high-risk systems under Annex I

Obligations that build on the high-risk requirements but sit outside Chapter III Sections 1 to 3 (conformity assessment, registration, post-market monitoring, incident reporting) have nothing to attach to without those requirements. Cenedril therefore dates them to the application of the requirements they relate to.

Standards for the AI Act

Only harmonised standards whose reference is published in the Official Journal count for the presumption of conformity under Art. 40. With Implementing Decision (EU) 2026/2274 of 12 October 2026, so far this is EN 18286 for the quality management system of providers (Art. 17). Further standards for risk management, data, accuracy and cybersecurity are being developed by CEN-CENELEC JTC 21. ISO/IEC 42001 is not harmonised, but provides the management system in which you run the obligations on an ongoing basis.

Since the Digital Omnibus, conformity with the Cyber Resilience Act also gives a presumption for cybersecurity under Art. 15 (Art. 42(3)).

Relationship to other laws

  • GDPR: If an AI system processes personal data, the GDPR applies alongside. The fundamental rights impact assessment may refer to the data protection impact assessment (Art. 27(4)).
  • Cyber Resilience Act: Products with digital elements that contain AI fall under both regulations.
  • NIS2: For entities under NIS2, AI systems are part of the network and information systems whose security they manage.

Free templates

The AI Compliance Starter Kit maps the obligations of the Regulation to the roles and provides the templates for them: for everyone the AI policy and the AI usage policy, for deployers the notice to employees and the fundamental rights impact assessment, for providers the AI provider policy, the technical documentation, the EU declaration of conformity, the post-market monitoring plan and the serious incident report.

What Cenedril covers

Cenedril’s AI assistant reflects the regulation as amended by Regulation (EU) 2026/1744:

  • classification of every AI system: prohibited practices, high-risk under Annex I and III with the exception under Art. 6(3), transparency obligations,
  • determining the role and a catalogue of obligations per role with their date of application,
  • AI register, AI risk register and fundamental rights impact assessment,
  • technical documentation under Annex IV, conformity check and EU declaration of conformity,
  • post-market monitoring plan and review, incidents with the competent market surveillance authorities per country,
  • explanation requests under Art. 86 and complaints,
  • for providers, the quality management system under EN 18286.

Sources

Frequently asked questions

Does the AI Act affect me if I only buy and use AI?

Yes. Whoever uses an AI system under their own authority is a deployer. All deployers are bound by the prohibited practices (Art. 5) and the obligation to support the AI literacy of their staff (Art. 4). Deployers of high-risk systems also have their own obligations under Art. 26, such as human oversight, monitoring of operation and keeping the logs. Certain deployers must carry out a fundamental rights impact assessment before use (Art. 27).

When do the obligations for high-risk AI apply?

Regulation (EU) 2026/1744 (Digital Omnibus on AI) postponed the dates of application. The requirements for high-risk systems under Annex III apply from 2 December 2027, for high-risk systems under Annex I (products under EU harmonisation legislation) from 2 August 2028. Prohibitions and AI literacy have applied since 2 February 2025, the obligations for general-purpose AI models since 2 August 2025.

What fines apply?

For prohibited practices up to EUR 35 million or 7% of worldwide annual turnover, for infringements of most other obligations up to EUR 15 million or 3%, for incorrect or incomplete information to authorities up to EUR 7.5 million or 1%. For small and medium-sized enterprises, the lower of the two amounts applies in each case (Art. 99).

No Cookies!

This wiki collects nothing, bakes nothing and leaves nothing behind. There's nothing to consent to. Privacy doesn't get better than this.