The short answer: a US company is in scope when any one of three triggers is met — the system is placed on the EU market, it is put into service or used in the EU, or its output is used in the EU. No EU office, EU servers or EU registration is required for the Act to reach you. What you then owe depends on your role (provider or deployer) and your system's risk category — in that order.
The 3 scope triggers, in plain English
The Act's scope rule (Article 2) is written for lawyers. Reduced to the questions a US business actually has to answer, it comes down to three triggers. One is enough.
You place the system on the EU market Provider route
You offer the AI system to customers or users in the EU — paid or free, direct or through a distributor, as a product or inside a product. Selling a SaaS subscription to EU businesses, onboarding EU users, or marketing the system for EU use all point the same way: the system is being made available on the EU market, and the company that supplies it holds provider duties for it.
The system is used in the EU Deployer route
Your company — or your EU subsidiary, branch or team — uses an AI system in the EU: screening EU candidates, pricing for EU customers, supporting EU users, evaluating EU workers. You did not build the system, and that does not matter. The business that deploys it in the EU carries deployer duties for how it is used there.
The system's output is used in the EU Output route
The system runs outside the EU, but what it produces is used inside it — a score, ranking, decision, report or piece of content that an EU business or person acts on. The Act was written to close exactly this gap: moving the server does not move the effect, and it is the effect the Act follows.
What does not decide scope: where the company is incorporated, where the servers sit, where the model was trained, or what the vendor calls the product. Scope follows the market, the use and the output — nothing else on that list.
Six situations US companies actually ask about
Illustrative situations, written for this guide — not client cases and not legal rulings on any real company. “Likely” means the trigger analysis points that way on these facts; a specific company on specific facts needs its own written assessment.
| Situation | Likely in scope? | Why |
|---|---|---|
| A US SaaS company sells an AI support agent to EU businesses | Likely yes | Trigger 1: the system is placed on the EU market. Provider duties follow the risk category. |
| A US company uses a vendor's AI tool to screen job applicants in the EU | Likely yes | Trigger 2: deployment in the EU — and hiring is an Annex III high-risk area, whoever built the tool. |
| A US internal tool is used only by a US team, on US customers, with no EU marketing or use | Likely no — keep checking | No trigger is met on these facts. The answer changes the moment an EU team, customer or output enters the picture. |
| A free AI tool is actively promoted to EU users, who sign up and use it | Can be yes | “Free” is not an exemption. Actively making a system available in the EU can be placing it on the market. |
| A US firm white-labels a general-purpose model and sells it under its own brand in the EU | Likely yes — as provider | Putting your name on a system and placing it on the EU market can move provider duties to you, on top of the model provider's own GPAI track. |
| A US company's website can simply be visited from Europe, with no EU offer, users or contracts | Grey — do not assume either way | Mere reachability is a weaker fact than an EU offer, onboarding or use. Record what you actually do in the EU market instead of guessing from traffic alone. |
Provider or deployer? Your role decides your duties
Provider — you develop the system, or have it developed, and place it on the market under your own name. Provider duties are the heavier set: for high-risk systems, risk management, data governance, technical documentation, logging, human-oversight design and a conformity assessment before market placement. Providers established outside the EU that place certain high-risk systems or general-purpose models on the EU market generally also need an authorised representative established in the EU appointed before they do so.
Deployer — you use a system someone else built, under your own authority, in your business. Deployer duties are narrower but they are yours: use the system as instructed, keep human oversight real rather than nominal, monitor its operation, keep the logs it generates, and — for the transparency layer — tell people when they are interacting with AI or seeing AI-generated content. A vendor contract can allocate tasks; it cannot hand your legal role to the vendor.
Both, sometimes. Substantially modify a system, change its intended purpose, or rebrand it as your own, and deployer duties can grow into provider duties. US companies hit this most often with white-labelled tools and heavily customised agents.
Scope is only step one: category and dates come next
Being in scope does not mean every duty applies to you. After scope, the order is always the same: role first, then risk category (prohibited, high-risk, limited or minimal, with the general-purpose model track alongside), then the dates for that category. Banned practices applied from February 2025 and GPAI duties from August 2025. Article 50 transparency duties applied from 2 August 2026; on 2 December 2026 the marking grace period for generative systems already on the market ends and two new Article 5 bans take effect. Under the Digital Omnibus (Regulation (EU) 2026/1744), high-risk duties apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems (information as of November 2026). Fines reach up to €35 million or 7% of global turnover for prohibited practices, and up to €15 million or 3% for most other violations.
Two free tools on this site continue from here: the Risk Classification Self-Check works out the likely category in the browser, and the Deadline Planner turns a category into a dated obligation timeline. Neither sends or stores anything.
5 mistakes US companies make about scope
- “We have no EU entity, so EU law cannot reach us.” Scope follows market, use and output — an EU entity was never the test.
- “Our vendor says the tool is compliant.” A vendor can be compliant as a provider while you are still exposed as a deployer. The roles are separate, and so are the duties.
- “It is only an internal tool.” Internal tools that screen, rank, price or evaluate people stop being low-stakes the moment they touch an Annex III area — internal describes the seating plan, not the risk.
- “We already did GDPR, so we are covered.” GDPR governs personal data. The AI Act governs the AI system itself — its category, documentation, oversight and transparency. Overlapping paperwork is not the same obligation.
- “It is free / a side project / still in beta.” Price and project status are not scope tests. An actively offered free system in the EU can be placed on the market like a paid one.
9-point scope checklist for a US company
- ☐ List every AI system your company supplies or uses — including vendor tools and agents, not only products you built.
- ☐ For each system, answer the three triggers in writing: EU market? EU use? EU use of its output?
- ☐ Mark your role per system: provider, deployer, or both (check white-label and heavy customisation).
- ☐ Note who is affected in the EU: customers, candidates, workers, users, or people scored by the output.
- ☐ Run the category check per system — prohibited first, then high-risk, then transparency, then minimal.
- ☐ If you are a non-EU provider of a high-risk system or GPAI model, check the authorised-representative duty before EU placement.
- ☐ Write down the dates that attach to each category — including 2 December 2026, 2 December 2027 and 2 August 2028 where they apply.
- ☐ Keep the reasoning, not just the conclusion: one dated page per system that a partner, insurer or customer could read.
- ☐ Set a re-check trigger: new EU market, new use, new autonomy or a rule change re-opens the scope answer.
Get your scope and category settled in writing
The Classification Report classifies one AI system in writing: likely risk category with section-by-section reasoning, the obligations that follow for your role, deadlines and a step-by-step action plan in plain English — $49 one-time, delivered by email. The optional EU AI Act Watch is $19/month and is never required.
Get My Report — $49 Start With the Free QuestionnaireFree first steps: Risk Classification Guide · Transparency Obligations Guide
Frequently asked questions
Does the EU AI Act apply to a US company with no EU office?
It can. The Act follows the market and the use, not the company's registration. If you place an AI system on the EU market, put it into service in the EU, or the output of your system is used in the EU, the Act can apply even if your company, servers and staff are all in the United States. Having no EU office does not by itself take you out of scope.
What are the three scope triggers for a US company?
In plain English: (1) you offer or sell the AI system in the EU — placing it on the market; (2) you use the system in the EU as a deployer, for example through an EU team, subsidiary or customers who use it there; and (3) people in the EU use the system's output, even if the system itself runs outside the EU. Any one of the three can bring the relevant duties with it.
If I only use an AI tool built by someone else, am I still covered?
Yes, in the deployer role. Buying a vendor's system does not transfer your deployer duties to the vendor. If your business uses AI with EU customers, candidates, workers or users — for screening, pricing, eligibility, support or content — you keep the duties that attach to how you deploy it, even though the provider built the model.
Does a free tool with EU users count?
It can. Placing on the market under the Act is not limited to paid sales — making a system available in the EU, including free access you actively offer to EU users, can count. Incidental, untargeted web traffic is a greyer area than actively marketing, onboarding or contracting with EU users. The safer test is what you actually do in the EU market, not whether money changed hands.
Do US companies have to appoint someone in the EU?
Providers established outside the EU that place certain high-risk AI systems or general-purpose AI models on the EU market generally need an authorised representative established in the EU before they do so. Whether that duty applies to you depends on your role and your system's category — which is why role and category come before any appointment decision.
When do the duties apply?
Banned practices applied from February 2025 and general-purpose AI duties from August 2025. Article 50 transparency duties applied from 2 August 2026; on 2 December 2026 the marking grace period for generative systems already on the market ends and two new Article 5 bans take effect. Under the Digital Omnibus, high-risk duties apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems (information as of November 2026).
Is this guide legal advice?
No. This guide and the SkillQuesters classification report are educational information based on the public text of the EU AI Act (Regulation (EU) 2024/1689). They do not replace a lawyer, and scope for a specific company on specific facts is a legal judgement.
SkillQuesters is independent and is not affiliated with the European Commission. Primary sources: Regulation (EU) 2024/1689, in particular Article 2 (scope), and the European Commission's AI Act pages. Timeline information as of November 2026. This page is educational information, not legal advice.