What does the EU AI Act require of deployers?
A deployer is an organisation that uses an AI system under its authority in a professional capacity. Under Regulation (EU) 2024/1689, every deployer must take AI literacy measures and avoid prohibited practices. Deployers of high-risk systems must follow the instructions for use, assign human oversight, monitor, keep logs and inform affected people. Some must also assess fundamental rights impact and meet transparency duties.
This guide is general information, not legal advice. Whether a system is in scope, and how it is classified, is a determination for your organisation and its counsel.
Provider or deployer: start with the role
Article 3 defines the roles. A provider develops an AI system, or has one developed, and places it on the market or puts it into service under its own name or trademark. A deployer uses an AI system under its authority, except for personal, non-professional use. Most obligations for high-risk systems fall on providers. Deployers carry a shorter, operational list.
The role can change. Under Article 25, a deployer is treated as the provider of a high-risk system if it puts its name or trademark on one, makes a substantial modification to one, or changes the intended purpose of a system so that it becomes high-risk. Record the role per system, not once for the whole organisation.
The timeline that matters to deployers
| Date | What applies | What it means for a deployer |
|---|---|---|
| 2 February 2025 | Definitions, AI literacy (Article 4), prohibited practices (Article 5) | Literacy measures in place and no prohibited use |
| 2 August 2025 | Obligations for providers of general-purpose AI models | Mostly provider duties; ask your vendors what they publish |
| 2 August 2026 | Transparency obligations (Article 50) and most other rules | Disclosures for emotion recognition, biometric categorisation and deep fakes |
| 2 December 2027 | High-risk rules for Annex III systems | Article 26 duties and, where it applies, the Article 27 assessment |
| 2 August 2028 | High-risk rules for AI in products covered by Annex I | Same duties for AI embedded in regulated products |
The Annex III and Annex I dates are those set by the Digital Omnibus on AI, Regulation (EU) 2026/1744, which amended the original timetable. The same amendment added two prohibited practices to Article 5, on non-consensual intimate imagery and child sexual abuse material. Check the consolidated text for the date from which they apply.
AI literacy (Article 4)
Article 4 applies to every provider and deployer, whatever the risk level. The Digital Omnibus on AI amended this provision; check the consolidated text. As amended, it asks providers and deployers to take measures to support the development of AI literacy of their staff and of other persons who operate or use AI systems on their behalf. It does not require a specific level of AI literacy from any individual. Keep a record of who received what, and why that was proportionate.
Prohibited practices (Article 5) as a check
Before anything else, confirm that no system in your inventory falls under Article 5. The practices include:
- Manipulative or deceptive techniques, and exploitation of vulnerabilities, that materially distort behaviour and cause significant harm
- Social scoring that leads to unjustified or unrelated detrimental treatment
- Predicting that a person will commit a crime based solely on profiling or personality traits
- Untargeted scraping of facial images to build facial recognition databases
- Emotion recognition in the workplace and in education, except for medical or safety reasons
- Biometric categorisation to infer sensitive characteristics such as political opinions or sexual orientation
- Real-time remote biometric identification in public spaces for law enforcement, outside narrow exceptions
Workplace emotion recognition is the item most likely to appear in an ordinary enterprise, inside an HR, call centre or meeting tool. Record the check and its result for each system.
Deployer obligations for high-risk systems (Article 26)
If a system is high-risk under Article 6 and Annex III or Annex I, Article 26 asks the deployer to:
- Use it according to the instructions for use, with appropriate technical and organisational measures.
- Assign human oversight to people with the necessary competence, training, authority and support.
- Check input data, where the deployer controls it, for relevance and representativeness for the intended purpose.
- Monitor operation, inform the provider of risks, suspend use where a risk arises, and report serious incidents.
- Keep the logs the system generates, where they are under the deployer's control, for at least six months unless other law says otherwise.
- Inform workers' representatives and affected workers before using a high-risk system at the workplace.
- Inform affected people when an Annex III system makes, or assists, decisions about them.
- Use the provider's information to carry out a data protection impact assessment under GDPR Article 35 where one is required.
- Cooperate with the competent authorities. Public authority deployers must also use only registered systems.
Fundamental rights impact assessment (Article 27)
Article 27 applies to deployers that are bodies governed by public law, private entities providing public services, and deployers of the Annex III systems for creditworthiness assessment and for life and health insurance risk assessment and pricing. Before first use, they must describe the processes in which the system is used, the period and frequency of use, the people likely to be affected, the specific risks of harm, the human oversight measures and the measures to take if risks materialise. The results are notified to the market surveillance authority. An existing DPIA can be reused where it already covers part of this.
Transparency obligations for deployers (Article 50)
Article 50 paragraphs 1 and 2 fall on providers. Paragraphs 3 and 4 fall on deployers and have applied since 2 August 2026:
- Emotion recognition and biometric categorisation: inform the people exposed to the system that it is operating.
- Deep fakes: disclose that image, audio or video content has been artificially generated or manipulated, with a lighter duty for evidently artistic, satirical or fictional work.
- Text published to inform the public: disclose that it was AI generated, unless it has had human review or editorial control and someone holds editorial responsibility.
The information must be clear and distinguishable, given at the latest at the first interaction or exposure.
A practical EU AI Act checklist
| Step | What to record | Owner | Where evidence lives |
|---|---|---|---|
| 1. Inventory every AI system | System, vendor, purpose, owner, users, data | AI governance lead | AI system inventory |
| 2. Record the role | Provider or deployer, and any Article 25 trigger | Legal | Inventory record, legal note |
| 3. Check Article 5 | Each prohibited practice considered, result | Legal with system owner | Classification record |
| 4. Classify | High-risk or not, Annex reference, reasoning | Legal and counsel | Classification record |
| 5. AI literacy | Audience, content, date, rationale | HR or learning lead | Training records |
| 6. Transparency | Article 50 trigger, notice text, where shown | System owner | Notice copy, screenshots |
| 7. Human oversight | Named people, training, authority to stop | System owner | Oversight assignment |
| 8. Logs and monitoring | Log location, retention of at least six months, incident route | Security or IT operations | Log store, runbook |
| 9. Workers and affected people | Who was informed, when, how | HR, system owner | Communications record |
| 10. FRIA and DPIA | Whether Article 27 applies, the assessment, notification | DPO with legal | Assessment file |
| 11. Review | Next review date, changes since last review | AI governance lead | Approval record, evidence log |
How MERIDVAR helps with the EU AI Act
MERIDVAR is self-hosted software that holds the record and the evidence. It does not certify compliance, and it does not decide your classification.
- Control mapping with computed status. EU AI Act controls are mapped, their status is computed from what is actually recorded, and controls that cannot be computed are labelled as such.
- Rules-based risk classification. Risk is classified as Low, Moderate, High or Critical by deterministic rules, with a reason for every factor and unknown never read as safe. The EU AI Act classification remains your organisation's determination.
- Evidence log. Decisions, reviews and approvals are written to an append-only, hash-chained log, with optional Ed25519 signatures.
- Audit package. A package per AI system or vendor collects the evidence and states what it does not prove.
See compliance for the frameworks covered and the product overview for the full path.
Sources: Regulation (EU) 2024/1689 on EUR-Lex; Regulation (EU) 2026/1744 (Digital Omnibus on AI) on EUR-Lex; AI Act Service Desk, implementation timeline; article texts at artificialintelligenceact.eu (Articles 3, 4, 5, 25, 26, 27 and 50).