Checklist for deployers of high-risk AI: your duties under Article 26 of the EU AI Act

What must a deployer of a high-risk AI system do? Article 26 of the EU AI Act lists the duties in twelve paragraphs. In short: you use the system according to the provider's instructions for use. You assign human oversight to suitable people. You monitor the operation, keep the logs for at least six months, and inform workers and the people affected. If you have reason to consider that the system presents a risk, you inform without undue delay and suspend its use. If you identify a serious incident, you inform the provider first.

For high-risk AI systems under Annex III, Article 26 applies from 2 December 2027. For systems under Annex I it applies from 2 August 2028 (Article 113, point (c), as amended by Regulation (EU) 2026/1744).

Below: all twelve paragraphs as a checklist, and the warning signs that oblige you to act. Then the reporting deadlines, and what the provider does compared with the deployer.

The quotes are from Regulation (EU) 2024/1689 in its consolidated version of 27 July 2026 (CELEX 02024R1689-20260727).

Who is a deployer?

Article 3, point (4), defines a deployer as "a natural or legal person, public authority, agency or other body using an AI system under its authority except where the AI system is used in the course of a personal non-professional activity".

An example: a company buys software that pre-sorts job applications and uses it in its HR team. The company is the deployer. The maker of the software is the provider. Such systems are listed in Annex III, point 4(a).

Under Article 3, point (3), the provider is whoever develops an AI system or has it developed, "and places it on the market or puts the AI system into service under its own name or trademark". A company can be both. If it develops a system and uses it itself, it is provider and deployer of that system (Article 3, points (3) and (11)).

From when does Article 26 apply?

Article 26 is in Chapter III, Section 3. Article 113, point (c), sets the dates. That section applies from 2 December 2027 to AI systems "classified as high-risk pursuant to Article 6(2) and Annex III". It applies from 2 August 2028 to systems classified as high-risk pursuant to Article 6(1) and Annex I. For systems placed on the market or put into service before those dates, Article 111(2) applies. The Regulation covers them only after a significant change in their design. Systems intended to be used by public authorities must comply by 2 August 2030.

Many pages still give 2 August 2026. That was the date for Annex III systems in the original 2024 text. Regulation (EU) 2026/1744 moved it. Article 113 now names fixed dates. They do not depend on harmonised standards being ready.

Some duties already apply today. Since 2 February 2025: AI literacy (Article 4) and most prohibited practices (Article 5). Since 2 August 2026: the transparency duties of deployers under Article 50(3) and (4). Articles 72 and 73 are in Chapter IX, not in Chapter III. Article 113 makes no exception for Chapter IX, so the general date applies there: 2 August 2026. In practice they reach a system only once it is high-risk under Article 6, so from the dates above.

The checklist: all twelve paragraphs of Article 26

Each row gives a paragraph, the duty and evidence you can show in an inspection. The evidence column holds our suggestions, not requirements of the Act.

Paragraph 3 creates no extra duty, and paragraph 10 concerns only post-remote biometric identification, mainly in criminal investigations. They are in the list so that no paragraph is missing.

Article 49(3) adds the detail for paragraph 8. It covers deployers that are public authorities or Union bodies, and persons acting on their behalf. They must register themselves, select the system and register its use in the EU database. This applies before they put into service or use a system listed in Annex III (except point 2).

Red flags: when you must suspend use or report

"Red flag checklist" is not a term of the Act. It usually means a list of warning signs. Article 26 names five situations in which you must act at once:

• You have reason to consider that using the system according to the instructions may result in a risk within the meaning of Article 79(1). You must, "without undue delay, inform the provider or distributor and the relevant market surveillance authority", and suspend the use of the system (paragraph 5).
• You identify a serious incident. You "immediately inform first the provider, and then the importer or distributor and the relevant market surveillance authorities" (paragraph 5).
• You cannot reach the provider. Then Article 73 applies "mutatis mutandis" (paragraph 5). Our reading: the reporting duty and the deadlines below then fall on you.
• You are a public authority or a Union institution, body, office or agency (paragraph 8). The system you plan to use is not registered in the EU database under Article 71. You do not use it, and you inform the provider or the distributor (paragraph 8).
• For post-remote biometric identification, the authorisation is rejected. You stop the use with immediate effect and delete the personal data linked to it (paragraph 10).

Write down for each situation who decides, who reports and to whom. When it happens, there is no time to work it out.

What is a serious incident, and which deadlines apply?

Article 3, point (49), names four cases. A serious incident is an incident or malfunctioning of an AI system that "directly or indirectly" leads to any of the following:

• "the death of a person, or serious harm to a person's health";
• "a serious and irreversible disruption of the management or operation of critical infrastructure";
• "the infringement of obligations under Union law intended to protect fundamental rights";
• "serious harm to property or the environment".

The report under Article 73 is made by the provider (paragraph 1). The deployer, under Article 26(5), informs first the provider and then the importer or distributor and the market surveillance authorities. The deadlines run from when the provider "or, where applicable, the deployer" becomes aware of the incident.

"At the latest" does not mean "wait until then". The report is due immediately once the causal link, or its reasonable likelihood, is established. In case of death, it is due as soon as the link is suspected.

Monitoring: Article 26(5) is not Article 72

The deployer monitors the operation of its system "on the basis of the instructions for use" and, where relevant, informs the provider (Article 26(5)).

The provider establishes and documents a post-market monitoring system (Article 72(1)). It collects and analyses data on performance over the whole lifetime, including data provided by deployers (paragraph 2). The plan for it is part of the Annex IV technical documentation (paragraph 3).

In short: the deployer watches its own use and reports. The provider watches the product across all its customers and analyses. What you report as a deployer is part of the data the provider analyses under Article 72.

More in our article on continuous monitoring of AI systems.

Provider or deployer: which tasks must the AI provider do?

The provider's duties are in Article 16, points (a) to (l). The deployer's duties are in Article 26. The table puts them side by side. Monitoring and incident reporting are in Articles 72 and 73; the quality management system of Article 17 covers them.

A common quiz question asks which tasks must be done by the AI provider. The answer is in Article 16. The provider meets the requirements and runs the quality management system. It also does the conformity assessment, the EU declaration of conformity, the CE marking, the registration and the corrective actions. Using the system according to the instructions, human oversight in daily use, monitoring the operation and informing workers are the deployer's tasks.

When do you become the provider yourself?

Under Article 25(1), you are considered a provider and take on the obligations of Article 16 if you:

• put your name or trademark on a high-risk AI system already placed on the market or put into service (point (a)). Contracts may still allocate the obligations otherwise;
• make a substantial modification to such a system and it remains high-risk (point (b));
• modify the intended purpose of a system not classified as high-risk, including a general-purpose AI system, so that it becomes high-risk (point (c)).

An example for point (c): you use a general-purpose chat assistant to rate job applications. The system can then fall under Annex III, point 4(a), and for that use you can become the provider.

What that means for the technical documentation: Annex IV template and Article 11.

Duties next to Article 26

AI literacy (Article 4). Providers and deployers "shall take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf". A literacy duty has applied since 2 February 2025 to all AI systems; this wording comes from Regulation (EU) 2026/1744. It sets no level: the Act does not require deployers "to guarantee any specific level of AI literacy of any individual".

To record your measures, use our free AI literacy training register.

Fundamental rights impact assessment (Article 27). Required before deploying a high-risk AI system under Article 6(2) (Annex III). It applies to bodies governed by public law and to private entities providing public services. It also applies to deployers of systems under Annex III, points 5(b) and 5(c). These are creditworthiness and credit scoring, and risk assessment and pricing in life and health insurance. Systems in Annex III, point 2 are excluded.

Who supervises in Germany?

The German act on AI market surveillance and innovation (KI-MIG) entered into force on 29 July 2026. It makes the Bundesnetzagentur the market surveillance authority, contact point and complaints office for the EU AI Act.

The Bundesnetzagentur's AI Service Desk gives guidance and an online tool for a first risk-class check. Under Article 74(3), (6) and (8), other authorities supervise some systems, for example the financial supervisor for systems used by financial institutions.

How to start

1. List your AI systems and check for each one whether it falls under Annex III. Our free self-assessment is a first step.
2. For each high-risk system, ask the provider for the instructions for use. Without them you cannot meet paragraphs 1, 5 and 9.
3. Name the oversight persons and record their training.
4. Set the reporting route: who informs whom about a risk or an incident, with the contact details of the provider, the distributor and the authority.
5. Set the retention period for the logs: at least six months, unless other Union or national law provides otherwise. Align it with your deletion rules.

What LandingRed offers for this

In LandingRed, the «Deployer Obligations» page lists each AI system you deploy. For each one it shows eight of the twelve paragraphs of Article 26 and the fundamental rights impact assessment under Article 27. For each duty you see whether evidence is present, partial or missing. The provider's duties under Article 16 have their own page.

More on the EU AI Act in LandingRed, or start with the free self-assessment.

Frequently asked questions

What is a red flag checklist for AI?

The term is not in the EU AI Act. It usually means a list of warning signs. For deployers of high-risk AI, the main warning signs come from Article 26(5). A risk within the meaning of Article 79(1): inform and suspend use. A serious incident: inform the provider first.

From when do the deployer duties of Article 26 apply?

For high-risk AI systems under Annex III, from 2 December 2027. For systems under Annex I, from 2 August 2028 (Article 113, point (c), as amended by Regulation (EU) 2026/1744). For systems placed on the market earlier, Article 111(2) applies. The AI literacy duty of Article 4 has applied since 2 February 2025.

Must the deployer report serious incidents to the authority itself?

The deployer informs the authority, but the formal report under Article 73 is made by the provider. Under Article 26(5), the deployer immediately informs first the provider, and then the importer or distributor and the market surveillance authorities. If the deployer cannot reach the provider, Article 73 applies to it mutatis mutandis.

Which tasks must the AI provider do, not the deployer?

The tasks of Article 16. The provider meets the requirements for the system and runs the quality management system. It carries out the conformity assessment, draws up the EU declaration of conformity and affixes the CE marking. It registers the system under Article 49(1), takes the corrective actions and ensures accessibility. The deployer uses the system according to the instructions, oversees it, monitors the operation and keeps the logs.

How long must the deployer keep the logs?

At least six months, to the extent the logs are under its control and other Union or national law does not provide otherwise (Article 26(6)). The period must be appropriate to the intended purpose of the system. Financial institutions subject to internal-governance requirements under Union financial services law keep them as part of the documentation required by that law.

Related resources

• Continuous monitoring of AI systems: Article 72 and Article 26(5)
• Technical documentation under Article 11: who draws it up after a role change
• Annex III high-risk classification: is your system high-risk?
• EU AI Act compliance checklist
• LandingRed for the EU AI Act: inventory, classification, deployer duties

This page gives information, not legal advice. The text quoted is Regulation (EU) 2024/1689 in its consolidated version of 27 July 2026 (CELEX 02024R1689-20260727). The information on Germany is current as of 2 November 2026. Only the texts published in the Official Journal of the European Union are authentic.

Reviewed and published under the editorial responsibility of AB Corporate Advisory S.R.L.