# AI Act deployer obligations: what Luxembourg companies using AI must do, and when

> Companies that use AI in Luxembourg are deployers under the AI Act: literacy, bans and transparency apply now, high-risk duties from 2 December 2027.

Published: 2026-11-17 · Dataplaz editorial team · https://dataplaz.lu/en/ai-act-deployer-obligations-luxembourg/

## Key takeaways
- Under Article 3(4) of the AI Act, a deployer is any organisation that uses an AI system under its authority, except for purely personal, non-professional use.
- AI literacy (Article 4) and the prohibited practices (Article 5) have applied to deployers since 2 February 2025; the Article 50 transparency duties for deepfakes and emotion recognition since 2 August 2026.
- The obligations for deployers of high-risk AI systems in Article 26 apply from 2 December 2027 for systems listed in Annex III and from 2 August 2028 for AI in Annex I products, as set by Regulation (EU) 2026/1744.
- Breaches of Article 26 or Article 50 can be fined up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher; for SMEs the lower amount applies.
- In Luxembourg, draft law No. 8476 would make the CNPD the default supervisor, with the CSSF and the Commissariat aux assurances for the entities they supervise; on 17 November 2026 the bill was still in committee.

Any Luxembourg company or public body that uses an AI system for professional purposes is a "deployer" under the EU AI Act, Regulation (EU) 2024/1689. Some deployer obligations already apply: AI literacy and the prohibited practices since 2 February 2025, and transparency duties for deepfakes and emotion recognition since 2 August 2026. The twelve obligations for deployers of high-risk AI systems in Article 26 apply from 2 December 2027 for systems listed in Annex III, after the Digital Omnibus, Regulation (EU) 2026/1744, postponed them.

## Who counts as a deployer

Article 3(4) defines a deployer as a natural or legal person, public authority, agency or other body "using an AI system under its authority", except in the course of a personal, non-professional activity. A bank running a credit-scoring model bought from a vendor is a deployer. So is a municipality with a chatbot, or an HR team that uses software to rank job applications.

The provider is the one who develops the system or places it on the market under its own name. The split matters, because the AI Act puts most of the technical burden on providers and most of the burden of *use* on deployers.

**Buying an AI tool does not buy you out of the AI Act.**

## What applies to every deployer now

Consider the questions a supervisor could ask today. Who in your organisation knows which AI tools staff use? Who decided they may be used, and for what?

| From | Obligation | Applies to |
|---|---|---|
| 2 February 2025 | AI literacy (Article 4) | All providers and deployers |
| 2 February 2025 | Prohibited practices (Article 5) | All operators |
| 2 August 2026 | Inform people exposed to emotion recognition or biometric categorisation (Article 50(3)) | Deployers of those systems |
| 2 August 2026 | Disclose deepfakes and AI-generated text published to inform the public (Article 50(4)) | Deployers of those systems |
| 2 December 2026 | New bans on non-consensual intimate imagery and child sexual abuse material | Deployers only if they use the system to generate such material |

Article 4 was rewritten on 27 July 2026: deployers must now "take measures to support" the AI literacy of their staff, and the text says this does not require a guaranteed level for any individual. The text-disclosure duty in Article 50(4) does not apply where AI-generated content has undergone human review or editorial control and someone holds editorial responsibility.

## The twelve obligations for high-risk deployers

A system is high-risk if it is listed in Annex III, such as AI for recruitment, creditworthiness assessment of natural persons, or risk assessment and pricing in life and health insurance, or if it is a safety component of a product covered by Annex I. For these systems, Article 26 requires deployers to:

1. use the system in line with the provider's instructions for use, through technical and organisational measures;
2. assign human oversight to people with the necessary competence, training, authority and support;
3. ensure that input data they control is relevant and sufficiently representative for the intended purpose;
4. monitor the system, and inform the provider or distributor and the market surveillance authority and suspend use if it presents a risk;
5. report serious incidents, first to the provider, then to the importer or distributor and the authorities;
6. keep the automatically generated logs under their control for at least six months;
7. inform workers' representatives and affected workers before using the system at the workplace;
8. for public authorities: register under Article 49 and not use an unregistered system;
9. use the provider's information to carry out a GDPR data protection impact assessment where one is required;
10. for post-remote biometric identification in criminal investigations: obtain judicial or administrative authorisation;
11. inform natural persons that they are subject to an Annex III system that makes or assists decisions about them;
12. cooperate with the competent authorities.

Two further duties sit outside Article 26. Public bodies, private entities providing public services, and deployers of credit-scoring and life and health insurance systems must carry out a fundamental rights impact assessment (Article 27) before first use and notify the results to the market surveillance authority. Since the Omnibus, they may cross-reference the relevant sections of an existing GDPR impact assessment. And under Article 86, a person affected by a decision based on most Annex III systems can ask the deployer for a clear and meaningful explanation of the AI's role in that decision. More detail is in the [explainer on the DPIA and the FRIA](/en/dpia-fria-ai-systems-luxembourg/).

## The timeline, and the grace period

The Digital Omnibus moved Article 26 to **2 December 2027** for Annex III systems and to **2 August 2028** for AI in Annex I products. It also clarified that high-risk systems already on the market before those dates are only caught if their design changes significantly afterwards. Systems intended to be used by public authorities must comply by 2 August 2030 in any case. The [analysis of the Omnibus delay](/en/ai-act-high-risk-obligations-delay-2026/) has the full timeline.

## When a deployer becomes a provider

Article 25 is where many deployers get caught out. A deployer is treated as the provider of a high-risk system if it puts its own name or trademark on it, makes a substantial modification, or changes the intended purpose of a system so that it becomes high-risk. Take a general-purpose assistant that a team configures to filter job applications: that use falls under Annex III, and the full provider obligations could follow.

**Your intended use is a compliance decision, not an IT setting.**

## What it means in Luxembourg

The financial sector gets a specific concession. For financial institutions such as banks that are subject to internal governance rules under EU financial services law, the monitoring duty is deemed fulfilled by complying with those governance rules, and logs are kept as part of the documentation that law already requires.

Under Article 7 of [draft law No. 8476](https://www.chd.lu/fr/dossier/8476), the Commission nationale pour la protection des données (CNPD) would be the default market surveillance authority, the Commission de Surveillance du Secteur Financier (CSSF) would supervise AI used by the entities it supervises, and the Commissariat aux assurances (CAA) would cover insurers. On 17 November 2026 the parliamentary dossier still showed the bill in committee, with no update since the Conseil d'État opinion of 10 July 2026. The [overview of AI Act supervisors in Luxembourg](/en/ai-act-luxembourg-authorities-cnpd/) explains the split.

The stakes are real. Breaches of Article 26 or Article 50 can be fined up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher, and for SMEs whichever is lower. Until the national law is voted, Luxembourg's own penalty rules are not in place.

The good news: most of this is organisation, not technology. A deployer needs to know what it uses, who owns each use and what the provider promised, and most of that is a list, a name and a contract clause.

## What to do now

1. **Build the AI inventory.** List every AI system in use, including cloud and generative tools, and record for each whether you are provider or deployer.
2. **Classify each use.** Check each one against Annex III and Article 5, and flag any use that changes a tool's intended purpose.
3. **Name an owner per system.** One person who can make the call on oversight and suspension, not a committee. For the oversight and decision structure, [an independent review of AI governance readiness](https://harro.lu/work-with-me) can help set priorities.
4. **Read the contracts.** Make sure providers deliver instructions for use, log access and incident contacts, because Article 26 depends on them.
5. **Prepare people and paperwork.** Plan AI literacy measures, workers' information and, where required, the FRIA before 2 December 2027.

## Questions & answers
**Is my company a provider or a deployer under the AI Act?**
If you develop an AI system or place it on the market under your own name, you are a provider. If you use an AI system built by someone else for professional purposes, you are a deployer. Most Luxembourg companies using AI tools are deployers.

**Can a deployer become a provider?**
Yes. Under Article 25, a deployer is treated as the provider of a high-risk AI system if it puts its name or trademark on it, makes a substantial modification to it, or changes the intended purpose of a system so that it becomes high-risk.

**Do deployers of high-risk AI systems have to register them?**
Only deployers that are public authorities or EU bodies must comply with the registration obligations of Article 49. They may not use a high-risk system that is not registered in the EU database.

**How long must a deployer keep the logs of a high-risk AI system?**
For a period appropriate to the intended purpose, and at least six months, unless other EU or national law, in particular data protection law, provides otherwise. Financial institutions keep them as part of the documentation required by financial services law.

## Sources
1. [Regulation (EU) 2024/1689 (AI Act)](https://eur-lex.europa.eu/eli/reg/2024/1689/oj) · EUR-Lex
2. [Regulation (EU) 2026/1744 (Digital Omnibus on AI)](https://eur-lex.europa.eu/eli/reg/2026/1744/oj) · EUR-Lex
3. [Projet de loi 8476 – dossier parlementaire](https://www.chd.lu/fr/dossier/8476) · Chambre des Députés
4. [Projet de loi n° 8476 – document de dépôt (23 December 2024)](https://wdocs-pub.chd.lu/docs/exped/0151/094/302942.pdf) · Chambre des Députés
5. [Article 26: Obligations of deployers of high-risk AI systems](https://artificialintelligenceact.eu/article/26/) · artificialintelligenceact.eu
6. [Digital Omnibus on AI](https://artificialintelligenceact.eu/ai-act-explorer/digital-omnibus/) · artificialintelligenceact.eu
