EU AI Act for companies: what it requires, when it applies and how to prepare
Regulation (EU) 2024/1689 — the EU AI Act — is the European Union's first general law on artificial intelligence. It does not regulate “AI” in the abstract: it assigns specific obligations according to the role your company plays with respect to each system and to the risk of each use. It has been applying in phases since 2 February 2025, and it has been amended by Regulation (EU) 2026/1744, in force since 27 July 2026. This guide covers the essentials for a company that uses or procures AI.
What it is — and what it is not
The Regulation classifies AI uses by level of risk and distributes obligations among the actors in the chain: whoever develops and markets a system, whoever imports or distributes it, and whoever uses it in the course of business. It is not a data protection law — that remains the GDPR, which applies in parallel — nor does it require you to “certify” your company's AI: it requires, depending on the case, documenting, informing, overseeing and being able to demonstrate what you do.
Where to start: by role, by date and by sector
This guide summarises; the detail is in articles that each answer one question. If you do not know where to start, this is a reasonable order: four questions, and for each one the reading that answers it.
1. Does it apply to me?
Before classifying anything, Article 2 sets out whom the Regulation reaches, and the inventory shows which AI the company actually uses, including the AI that arrived without a contract. If your doubt is about your line of business, the sector cases point to where high risk usually sits.
- Does the AI Act even apply to you? Article 2 answers before any classification
- The AI you're already using without knowing it: why the inventory is the foundation of everything
- Minimal risk or high risk: what each one requires, article by article
- Your company's high risk probably isn't in your business: it's in HR
By sector:
- Recruitment: when HR is the business, high risk stops hiding
- Financial services: high risk is in your business, but narrower than you'd think
- Private education: Annex III has four letters for your school, and the fourth surprises everyone
- Private healthcare: the sector where high risk really is in the core activity — and through two different routes
- Advisory firms and law practices: the Annex III point you fear isn't about you
- All sectors
2. What am I in relation to each system?
Role goes system by system: most companies deploy, and some also provide. Article 25(1) sets out when a deployer is considered a provider, and all three circumstances concern high-risk systems. The quality management system of Article 17 belongs to the provider.
- When you stop being the deployer and become the provider: the three scenarios under Article 25
- You're fine-tuning an open model: have you become a provider?
- The Article 17 quality management system isn't yours to set up
- Article 26, paragraph by paragraph: what binds the deployer of high-risk AI, and which post covers each
3. Which date applies to me?
Each obligation has its own date of application, and Regulation (EU) 2026/1744 moved several of them. The summary is in this guide's dates, and the detail, article by article, in the calendar.
- Annex III high-risk applies from 2 December 2027
- What the Digital Omnibus changes in the AI Act
- The article that didn't move: why 2 August is still a real date
- The two new Article 5 prohibitions aren't 'category 9'
- AI medical devices: what changes on 2 August 2028 for manufacturers and for the clinics that use them
4. Which document do I need?
It depends on the previous answers, and the GDPR applies in parallel, without waiting for the AI Act timeline. The AI literacy measures of Article 4 already apply; the FRIA and the DPIA are different instruments, each with its own legal basis; and a complete SME case shows how a decision ends up documented.
- AI literacy: what Article 4 requires now that the Omnibus rewrote it
- The GDPR already applies to your AI: five articles that don't wait for 2027
- FRIA and DPIA are not the same thing: why one does not replace the other
- The AEPD's Article 35(4) list: when your AI requires a DPIA even if it is not high-risk under the AI Act
- A complete SME case: from the first AI inventoried to a reviewable decision (and one that does not apply)
Provider or deployer? The role goes system by system
The two central figures are the provider — whoever develops an AI system and places it on the market or puts it into service under its own name or trademark — and the deployer: whoever uses an AI system “under its authority”, except where the use is a personal, non-professional activity (Art. 3, point 4). That covers almost any company that procures AI-powered software.
The question “are we a provider or a deployer?” is badly framed: the Regulation assigns the role per system, not per company. Within the same inventory you can be the deployer of twelve tools and the provider of one.
And the role can change. Art. 25(1) sets out three circumstances in which a deployer comes to be considered a provider, and all three are confined to high risk: putting your name or trademark on a high-risk system already placed on the market; substantially modifying a high-risk system in such a way that it remains high-risk; or changing the intended purpose of a system that was not high-risk in such a way that it becomes so. Those three Article 25 scenarios are limited to high-risk AI systems. For other systems, placing your name or trademark on the system does not by itself trigger Article 25(1), but provider status must still be assessed against the general definition in Article 3(3).
Risk levels and transparency obligations
- Prohibited practices (Art. 5): banned uses regardless of who carries them out, applying since 2 February 2025.
- High risk (Art. 6 and Annexes I and III): permitted, with strict requirements for the provider and obligations of its own for the deployer.
- Transparency obligations (Art. 50): not a risk level, but obligations triggered by the nature of the system — chatbots and synthetic-content generation, among others — which require informing or marking, whatever the system's classification.
- All other AI systems: no specific obligations under the Regulation.
Classification is done system by system, with the rationale documented: it is the piece everything else rests on.
How to read an obligation: LM, LC, BP and IM
Alethexis tags every requirement with one of four labels. They are not categories of the Regulation: they are how we separate what binds you today from what will bind you later, and from what is only recommended.
- LM · Legally Mandatory.
- Legal obligation in force for your specific case. An LM obligation may depend on a factual condition (for example, only if the system recognises emotions): the condition does not make it a future obligation.
- LC · Legally Conditional.
- Legal obligation with a confirmed future date: it applies from the date of application shown in the platform.
- BP · Best Practice.
- International best practice (ISO, NIST, OECD) — never a direct legal obligation.
- IM · Internal Methodological.
- Alethexis’s own methodological control, with no direct basis in law.
The label depends on who you are in relation to each system and on the date: that is why the same article can bind one company and not appear on another's record. The four are Alethexis's own labels: no authority issues them.
Art. 4 — AI literacy, in force since February 2025
Providers and deployers must take AI literacy measures reaching their staff “and other persons dealing with the operation and use of AI systems on their behalf”. As amended by Regulation (EU) 2026/1744 it is expressly a measures-based obligation: the article itself says that it “does not require providers or deployers to guarantee any specific level of AI literacy of any individual”. You comply by documenting which measures you took and the basis on which you determined their scope — not by piling up diplomas.
Art. 5 — prohibited practices
The Art. 5 list has been applying since 2 February 2025: harmful subliminal manipulation, exploitation of vulnerabilities, social scoring, and certain uses of biometric identification, among others. Regulation (EU) 2026/1744 added two new prohibitions — Art. 5(1), points (ba) and (bb), on non-consensual intimate content and child sexual abuse material generated with AI — applicable from 2 December 2026. It is the part of the Regulation where risk is not managed: it is eliminated.
Art. 50 — transparency, applying since August 2026
Paragraphs 1, 3 and 4 of Art. 50 apply since 2 August 2026: informing people that they are interacting with an AI, and disclosing certain uses such as deepfakes. The information duty of Art. 50(1) is the provider's; for a deployer, reasonable diligence is verifying that its tools comply. The marking of synthetic content under Art. 50(2) also applies from 2 August 2026, with a transitional regime: for providers of systems placed on the market before that date, Art. 111(4) sets 2 December 2026 as the deadline.
High risk — Annex III, Art. 26 and the dates that matter
The obligations for high-risk systems under Annex III — including the deployer obligations of Art. 26: use in accordance with the instructions for use, human oversight entrusted to persons with the necessary competence, training and authority, monitoring and log-keeping — apply from 2 December 2027. For systems that are high-risk as regulated products under Annex I (Art. 6(1)), Art. 26 applies from 2 August 2028.
Two transitional rules worth knowing: high-risk systems already placed on the market before the date of application are only covered if they undergo significant changes in their designs (Art. 111(2)); and those deployed by public authorities before 2 December 2027 have 2 August 2030 as their horizon.
Annex III examples with direct impact on SMEs: recruitment and staff evaluation, access to essential services, credit scoring, education.
FRIA — the Art. 27 impact assessment, and who it actually binds
Before deploying certain high-risk systems under Annex III, some deployers must assess the impact on fundamental rights. The scope is narrower than it is usually told: it binds bodies governed by public law, private entities providing public services, and deployers of the systems in Annex III, point 5, letters (b) and (c). Its only date of application is 2 December 2027 — Art. 27 only reaches high-risk systems under Annex III, so the 2028 date for Annex I does not operate on it.
For most private companies, the FRIA will not be an obligation of their own. The honest thing is to say so — and it is still worth knowing what it is, because clients and tenders will mention it.
The FRIA is not the DPIA of Art. 35 GDPR: they are independent instruments, each with its own law. Since Regulation (EU) 2026/1744, Art. 27(4) expressly allows the FRIA to include cross-references to the relevant sections of the DPIA, or relevant parts of it, where an obligation is already met there — reusing material is enabled; the FRIA still has to be carried out, and conflating the instruments is not allowed.
General-purpose AI models (GPAI)
The obligations of the chapters on general-purpose models (Arts. 53–55) apply since 2 August 2025 and fall on the providers of those models. If your company uses them through commercial tools, your work is one of diligence: knowing which models sit behind your tools and what documentation each provider offers. Models already on the market before that date have a transitional regime until 2 August 2027 (Art. 111(3)).
The dates, at a glance
Dates, in order: Art. 4 and Art. 5 (2 February 2025) · GPAI and governance (2 August 2025) · Art. 50(1)/(3)/(4) and 50(2) (2 August 2026; for 50(2), Art. 111(4) sets 2 December 2026 as the deadline for providers of systems placed on the market before 2 August 2026) · the new prohibitions of Art. 5(1), points (ba) and (bb) (2 December 2026) · GPAI transitional period and sandboxes (2 August 2027) · Annex III, Art. 26 and Art. 27 (2 December 2027) · Annex I (2 August 2028) · pre-existing high-risk systems of public authorities (2 August 2030).
How to prepare: the Route in seven steps
Before the first step comes step 0: put in writing who is answerable for AI in your company, with their scope and their deputy.
- Inventory what is already in use — including what arrived without a contract. Inventory template →
- Assign the role per system: provider or deployer, row by row.
- Classify the risk of each use with the rationale in writing, its date and its author.
- Ask each vendor for its documentation and keep it in one file per vendor.
- Adopt and communicate the AI use policy, with a read acknowledgement from each person.
- Document your Art. 4 measures, already in force. AI literacy checklist →
- Keep traceable evidence of all of the above: it is what clients, auditors and authorities will ask you for, and a large client often asks for it before the legal deadline arrives.
The whole Route: one chapter per step, with its template and its document →
Frequently asked questions
What is the EU artificial intelligence regulation?
The EU artificial intelligence regulation is Regulation (EU) 2024/1689, known as the AI Act: the European Union’s first general law on artificial intelligence. It does not regulate “AI” in the abstract — it classifies uses by risk level and allocates obligations between whoever develops and places a system on the market, whoever imports or distributes it and whoever uses it in their activity. It has been amended by Regulation (EU) 2026/1744, in force since 27 July 2026.
What does it ask of a company that only uses third-party AI tools?
It depends on the system. Whoever uses an AI system under their own authority is the deployer (Art. 3, point 4), and that covers almost any company that buys software with AI in it. What the regulation asks of them, depending on the case, is to document, inform, oversee and be able to show what they do. It does not ask for a seal on your company’s AI, and it does not replace the GDPR, which applies in parallel.
Are we a provider or a deployer?
The question is put the wrong way round: the role is assigned per system, not per company. In the same inventory you can be the deployer of twelve tools and the provider of one. Art. 25(1) sets out three circumstances in which a deployer comes to be considered a provider, and all three are confined to high risk.
Since when does the AI Act apply?
In phases since 2 February 2025, when Art. 4 and Art. 5 became applicable. The remaining dates are in the “The dates, at a glance” section of this guide and, in detail, in the AI Act calendar.
Related reading
- AI governance software
- Shadow AI: the AI your company already uses without knowing it
- What the Digital Omnibus changes in the AI Act
- The EU's official icons for labelling AI content: how to use them and what they don't prove
- Who enforces the AI Act in Spain: AESIA and what a market surveillance action looks like
- Who must warn that your chatbot is an AI? The nuance in Article 50(1) that almost everyone gets wrong
- FRIA and DPIA are not the same thing: why one does not replace the other
- When FRIA and DPIA overlap: Article 27(4) lets you reuse, not skip one
- Article 26, paragraph by paragraph: what binds the deployer of high-risk AI, and which post covers each
- The GDPR already applies to your AI: five articles that don't wait for 2027
- Article 50 on one page: the four paragraphs, who each one binds, and the eight posts that unpack them
Dates under Art. 113 of Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744 (OJEU of 24 July 2026, in force since 27 July 2026). This content is informational and is not legal advice.