ProtForge

EU AI Act for US companies: what applies if you sell AI into Europe

The EU AI Act, Regulation (EU) 2024/1689, does not stop at the EU border. If you are a US SaaS or AI company with European customers, parts of it may apply to you even without an office in Europe. This is a plain-language map of what to check first, and a fixed-price remote scoping review if you want it done: $5,500, two weeks, up to five AI systems.

Book a scoping review Does it apply to us?

Updated: 21 Nov 2026. This is a practical guide, not legal advice.

Does the Act apply to you?

Article 2 sets an extraterritorial scope. The two cases that matter most for a US company:

You place an AI system or model on the EU market

Providers that place AI systems on the market or put them into service in the Union, or place general-purpose AI models on the market, are covered wherever they are established. Selling or offering your product to customers in the EU is the typical trigger.

Your output is used in the EU

Providers and deployers established or located in a third country are also covered where the output produced by the AI system is used in the Union. So an API or a service run from the US can be in scope if its results are used by people or companies in Europe.

What is outside

Systems that are not AI systems under the Article 3 definition, and uses outside the Union with no output used there, are not in scope. Scope is a question about each system and each use, not about your company as a whole.

Which role are you playing?

Obligations follow the role, and one company can hold several roles for different systems.

Provider

Develops an AI system, or has one developed, and places it on the market or puts it into service under its own name or trademark. Providers carry most of the obligations. If you build a product on top of a third-party model, you are normally the provider of the AI system, not of the model.

Deployer

Uses an AI system under its own authority in a professional context. Your EU customers will often be deployers of your product, and they will ask you for documentation and information in order to comply.

Importer and distributor

An importer places on the EU market a system bearing the name of a non-EU provider; a distributor makes it available further down the chain. If you sell through resellers in the EU, they may take these roles and have their own duties.

Authorised representative

Providers established outside the EU of high-risk AI systems must appoint an authorised representative in the Union before making the system available (Article 22). General-purpose model providers have a similar requirement under Article 54.

When a role changes

Under Article 25 a deployer or other party becomes the provider of a high-risk system if it puts its name or trademark on it, makes a substantial modification, or changes its intended purpose so that it becomes high-risk.

The risk tiers, in practice

The category follows the use of the system, so the same model can sit in different tiers.

Prohibited practices (Article 5)

Includes harmful manipulation, exploiting vulnerabilities, social scoring, emotion recognition in the workplace and in education (with limited exceptions), and untargeted scraping of facial images, among others. If a feature does one of these, it has to go.

High-risk (Annex III and Annex I)

Annex III lists areas such as employment and worker management, education, access to essential services including credit scoring, biometrics, and law enforcement. If you sell HR screening or credit tools, check this carefully. Article 6(3) allows some Annex III systems to document why they are not high-risk, but never where they profile natural persons.

Transparency (Article 50)

Chatbots must tell people they are interacting with an AI system unless it is obvious. Providers of systems that generate synthetic audio, image, video or text must mark the output in a machine-readable format, and deepfakes must be disclosed. This is the tier most SaaS products meet first.

General-purpose AI models

If you train a model and release it, provider obligations for general-purpose models apply, including technical documentation and a copyright policy, with more for models with systemic risk. If you only fine-tune, you become a provider of the modified model only where the change is substantial; the Commission guidance sets out how to assess that. Most teams that call an API or lightly fine-tune are not GPAI providers.

Everything else

Minimal risk systems have no specific obligations beyond AI literacy (Article 4), which applies to providers and deployers alike.

Dates and penalties

Prohibitions and AI literacy have applied since 2 February 2025. Obligations for general-purpose AI models have applied since 2 August 2025.

The timeline for high-risk systems has been the subject of EU simplification proposals, the so-called Digital Omnibus, so we do not give a specific high-risk date here: check the current text before you plan a roadmap. Classification is useful now regardless, because the tier does not change with the calendar.

Under Article 99, breaches of the prohibited practices can lead to fines of up to 35 million euros or 7 percent of worldwide annual turnover, whichever is higher. For SMEs, including startups, the lower of the two applies. Other infringements have lower ceilings, and fines are set case by case by the authorities.

A remote scoping review for US teams

We map your systems against the Regulation and tell you what applies, in writing, so your team and your lawyers can act on it.

Inventory and scope

A 30-minute call and a short questionnaire. We list up to five AI systems, who uses them, where, and whether Article 2 reaches you.

Roles and tiers

For each system: provider or deployer, risk tier with the reasoning written down, and any authorised representative requirement.

Obligations plan

A prioritised plan of what to do, what to document and what to watch as the dates are clarified. It is delivered at the end of week two.

The Act expects accuracy and robustness for high-risk systems, and customers increasingly ask for evidence of it. Our LLM evaluation audit, from $4,000, produces that evidence by measuring hallucination rates and failures on your own traffic, and RAG evaluation covers retrieval systems. The team comes from PhD-level research on LLM hallucination detection, and our comparison of hallucination detectors is public. We do not give legal advice; we work with your counsel.

FAQ

Does the EU AI Act apply to a US company with no EU office?

It can. Article 2 applies to providers that place AI systems or general-purpose AI models on the EU market, or put them into service in the EU, wherever they are established, and to providers and deployers in a third country where the output of the system is used in the Union. Having no EU entity does not take you out of scope; whether you are in scope depends on what the system does and where it is offered or used.

Do we need an EU authorised representative?

Providers established outside the EU of high-risk AI systems must appoint an authorised representative in the Union before making the system available there (Article 22). Providers of general-purpose AI models also have this requirement (Article 54), with some exceptions such as certain open-source models. If none of your systems is high-risk or a general-purpose model, the requirement does not apply to you.

Is our SaaS chatbot high-risk?

Usually not. A customer-facing chatbot normally falls under the transparency obligations of Article 50: people must be told they are interacting with an AI system unless it is obvious. It becomes high-risk only if its use is listed in Annex III, for example evaluating job candidates or creditworthiness. The classification follows the use, not the technology.

We fine-tune an open model. Does that make us a GPAI provider?

Not automatically. The obligations for providers of general-purpose AI models apply to whoever places the model on the market, and a downstream party that modifies a model becomes a provider only if the modification is substantial. The Commission has published guidance on this, so check it against what you actually do. Building a product on top of someone else's model usually makes you a provider of an AI system, not of the model.

When do the obligations apply?

The prohibitions and the AI literacy duty apply from 2 February 2025, and the obligations for general-purpose AI models from 2 August 2025. The timeline for high-risk systems has been the subject of EU simplification proposals, the so-called Digital Omnibus, so verify the current dates before you plan. We do not state a high-risk date here on purpose.

What are the penalties?

Under Article 99, violations of the prohibited practices can lead to fines of up to 35 million euros or 7 percent of worldwide annual turnover, whichever is higher; for SMEs including startups, whichever is lower. Other infringements have lower ceilings. Fines are set case by case and enforcement is by national authorities, and by the Commission for general-purpose models.

Is the scoping review legal advice?

No. It is a technical and regulatory mapping of your systems against the Regulation: roles, risk tier and the obligations that follow, with the reasoning written down. If your lawyers need to sign off on a conclusion, we work with them.

Tell us what you sell into Europe

A few lines on your product, your EU customers and what is on your roadmap. We reply within one business day and propose a 30-minute call. We work remotely with US teams and overlap with US East Coast mornings.

hello@protforge.com