Deployer — EU AI Act Definition and Obligations

Definition3 June 2026· 10 min read· 1,981 words

EU AI Act deployer: Article 3 definition, Article 26 duties, Article 27 FRIA scope, and when a deployer becomes a provider under Article 25.

Under Regulation (EU) 2024/1689, a deployer is any natural or legal person, public authority, agency or other body that uses an AI system under its authority — except where the system is used in the course of a personal, non-professional activity. The definition comes from Article 3, point 4. In practical terms, it covers every organisation that adopts a third-party AI tool and operates it for a business purpose: a retailer running a vendor's recommendation engine, a bank using a credit-scoring model built by a fintech provider, a logistics firm deploying a workforce-scheduling tool.

Deployer status matters because the EU AI Act attaches a distinct set of obligations to it — lighter than those of a provider, but real. And the status is not fixed: under Article 25, a deployer can cross into provider territory by modifying or rebranding a high-risk system.

The EU AI Act definition

Article 3, point 4 of Regulation (EU) 2024/1689 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."

Three elements carry weight. First, the deployer must be using the system — not building it, not placing it on the market for others, but operating it. Second, the use must be under its authority: the deployer controls whether, when and how the system runs. Third, the personal-use carve-out is narrow. A freelancer using a vendor's AI recruitment tool for client engagements is a deployer; an individual using a job-matching app for their own career search is not.

The deployer is distinct from the provider (Article 3, point 3), who develops or places an AI system on the market under their own name or trademark. Most organisations that buy or license AI tools — rather than building them — are deployers. That distinction shapes the compliance programme significantly.

What a deployer must do

Article 26 obligations

Article 26 sets out the core duties of deployers of high-risk AI systems. Read at article level; the paragraph numbering in secondary commentary is not always consistent across sources.

Use the system per the provider's instructions. The deployer must operate the AI system within the scope and purpose the provider has specified. Using a fraud-detection model for employment screening, for example, would be an out-of-scope deployment and could trigger Article 25 liability (see below).

Ensure human oversight. The deployer must assign the technical and organisational resources necessary to implement the human-oversight measures the provider has built into the system (Article 14 establishes human oversight as a requirement on providers; Article 26 makes deployers responsible for activating it). In practice this means designating competent staff who can interpret the system's outputs, override them where appropriate, and halt the system if needed.

Monitor operation and report upward. Deployers must monitor the system's performance, watch for risks, and — where a serious incident occurs — flag it to the provider. Serious-incident reporting to a national market-surveillance authority is the provider's duty under Article 73; the deployer's statutory obligation runs to the provider and, where relevant, to the authority.

Keep logs. Deployers must retain logs generated by the AI system for at least six months (Article 26). Do not confuse this with the technical documentation retention period (ten years, Article 18), which is a provider obligation.

Inform affected workers. Where the AI system is used in the workplace, deployers must notify employee representatives before deployment, respecting applicable national law.

Tell affected individuals. Deployers of high-risk AI systems that interact with natural persons must inform those persons that they are subject to the AI system's output. This obligation runs alongside the provider's transparency duties under Article 13.

Fundamental Rights Impact Assessment — Article 27

Not every deployer runs a Fundamental Rights Impact Assessment. Article 27 applies to:

Public bodies deploying high-risk AI systems.
Private entities providing public services — operators of critical infrastructure, providers of education or vocational training services, employment-service providers, providers of essential private and public services (including creditworthiness-scoring and life/health insurance), and others enumerated in Article 27's scope.

Private employers using an Annex III recruitment or performance-management tool are generally not required to run a FRIA. This is a common misconception. The FRIA obligation is narrower than it looks.

For those who must complete a FRIA, Article 27 requires a structured assessment of potential impacts on protected fundamental rights, taking account of the people affected, the context of use, and the nature of the high-risk system. Article 27 also allows the FRIA to build on an existing DPIA under GDPR Article 35 — the two assessments are distinct but compatible.

Article 50 transparency duties for limited-risk deployers

Deployers of limited-risk AI systems — chatbots, synthetic-media generators, emotion-recognition tools, deepfake producers — have transparency obligations under Article 50, which applies from 2 August 2026. The core duty is disclosure: people interacting with an AI system must be informed they are doing so. Deployers of synthetic audio, video, text or image content must label it as AI-generated or AI-manipulated, with specific carve-outs for artistic and satirical content.

Article 50 does not require a conformity assessment or technical documentation. It is a disclosure-only obligation.

When a deployer becomes a provider (Article 25)

A deployer does not permanently occupy a fixed role. Article 25 triggers a role shift — the deployer becomes a provider for the purposes of the Act — in three situations:

Rebranding. The deployer puts its own name or trademark on a high-risk AI system and makes it available to third parties or places it into service.
Substantial modification. The deployer makes a modification to a high-risk AI system that qualifies as substantial under Article 3, point 23. Substantial modification means a change that affects the system's conformity with the requirements of the Act, or alters its intended purpose to one not assessed by the original provider.
Change of intended purpose. The deployer modifies the intended purpose of an AI system — including a non-high-risk one — so that it becomes high-risk under Article 6.

The consequence of a role shift is serious. A deployer-turned-provider inherits the full high-risk provider stack: risk management under Article 9, technical documentation under Article 11, human oversight design under Article 14, a conformity assessment under Article 43, registration under Article 49, and the quality management system under Article 17, among others.

Non-compliance with the provider obligations carries a fine ceiling of €15,000,000 or 3% of total worldwide annual turnover — whichever is higher — under Article 99. For companies that qualify as SMEs or start-ups, Article 99(6) caps the fine at the lower of the percentage or the fixed sum.

Examples

A company using a third-party CV-screening tool

A 200-person staffing agency buys a vendor's AI-based CV-screening product to filter job applicants. The screening tool falls within Annex III, point 4(a) — recruitment and selection of natural persons — making it a high-risk AI system.

The agency is the deployer. Its Article 26 obligations: use the tool within the vendor's documented purpose; make sure trained staff review every output before a hiring decision is made; retain the system logs for at least six months; inform candidates they are subject to automated screening.

The agency is not required to run a FRIA (it is a private employer, not a public body or a listed service provider under Article 27). But if it decides to customise the screening model — adding a new scoring dimension not covered by the vendor's technical documentation — it should obtain legal advice on whether that constitutes a substantial modification triggering Article 25.

A bank deploying a vendor credit-scoring model

A regional lender integrates a fintech provider's creditworthiness-scoring engine into its loan application workflow. Creditworthiness AI falls under Annex III, point 5(b) — access to essential private services — and is high-risk.

The bank is the deployer. Article 26 requires it to operate the model within the vendor's intended purpose, assign human reviewers to assess borderline scores, keep a six-month log, and inform applicants. Because the bank is a provider of creditworthiness-assessment services — one of the types listed in Article 27 — it must also complete a Fundamental Rights Impact Assessment before deployment.

If the bank later builds its own internal credit model and rolls it out under its own brand, it becomes a provider for that system via Article 25, and the full high-risk compliance stack applies.

Frequently Asked Questions

Q: Is every company that uses an AI tool a deployer under the EU AI Act?

Only if the use is professional rather than personal or non-professional. Any business operating an AI system under its authority in the course of its commercial or public-sector activity is a deployer within the meaning of Article 3, point 4. The obligation level then depends on whether the system is high-risk, limited-risk, or minimal-risk — deployers of minimal-risk AI face no mandatory obligations under the Act.

Q: What is the penalty for a deployer that breaches Article 26?

Non-compliance with deployer obligations falls within the mid-tier of Article 99: up to €15,000,000 or 3% of total worldwide annual turnover, whichever is higher. For companies qualifying as SMEs or start-ups, Article 99(6) caps the fine at the lower figure. Penalties under Article 99 have applied since 2 August 2025.

Q: When do the high-risk deployer obligations actually kick in?

Under the Digital Omnibus (political agreement reached 7 May 2026), the application of the high-risk regime has been deferred. Stand-alone high-risk Annex III systems: 2 December 2027. High-risk AI embedded in Annex I regulated products: 2 August 2028. The original 2 August 2026 date no longer applies to high-risk obligations. Article 50 limited-risk transparency duties apply from 2 August 2026 on the original schedule.

Q: Does a deployer need to run a Fundamental Rights Impact Assessment?

Only certain deployers. Article 27 requires a FRIA from public bodies and from private entities providing the services listed in Article 27's scope — including providers of creditworthiness assessment and life/health insurance risk scoring. A private employer deploying a recruitment tool is generally not in scope for the FRIA. Check the Article 27 criteria against your specific context before concluding either way.

Q: Can a deployer become a provider?

Yes, under Article 25. Three triggers: rebranding the system under your own name; making a substantial modification (Article 3, point 23) to a high-risk system; or changing the intended purpose of any system such that it becomes high-risk. Each trigger attaches the full provider compliance stack to the organisation that crossed the line.

Q: A deployer uses an AI chatbot for customer service. Does Article 50 apply?

Yes, from 2 August 2026. Deployers of AI systems that interact with natural persons — including chatbots — must inform users that they are interacting with an AI, unless that is obvious from context. Article 50 also covers emotion-recognition systems, deepfake generators, and synthetic-content producers. Compliance here is a disclosure obligation only; no conformity assessment is required.

How Confir helps

Confir's classification workflow asks a structured set of plain-English questions about how your organisation uses each AI system. Its rule-based, deterministic engine derives your role — deployer, provider, importer, or distributor — and maps the obligations that follow. For deployers, it generates an Article 27 FRIA (where Article 27 applies), tracks the six-month log-retention requirement, and produces a Compliance Health Score across your AI inventory. Same intake, same finding, every time — no hallucination, no inference. Self-serve from €600/year at confir.eu.

Related terms

Manage your EU AI Act compliance in one place

Confir automates risk classification, technical documentation, and audit trails for any company. No consultants. No 6-month projects. 7-day free trial.

Start free trial →