In this article9 sections
The EU AI Act applies to eu ai act non eu companies just as firmly as it applies to Dutch or German businesses. If you ship an AI-powered product to a user in Amsterdam, Paris, or Warsaw, you are inside the scope of this regulation. Location of incorporation does not create an exemption.
This matters for US SaaS platforms, UK software vendors, Swiss HR tech providers, and anyone else whose output reaches people on European soil. The rules are not limited to companies with a legal entity in the EU. They follow the AI system to wherever its effects land.
What Article 2 Actually Says
Article 2 of the EU AI Act sets the territorial scope. The regulation applies to:
-
Providers who place an AI system on the EU market or put it into service in the EU, regardless of where the provider is established.
-
Deployers of AI systems who are located in the EU.
-
Providers and deployers located outside the EU when the output of the AI system is used inside the EU.
That last bullet is the critical one. A US provider whose AI system produces output used in the EU is in scope, even if it never sells to an EU entity directly. A Swiss platform whose outputs are consumed by EU-based HR departments falls in scope. The physical location of your servers is irrelevant. What matters is where the effect lands.
There are limited carve-outs, such as AI systems used exclusively for military, defence or national security purposes, and AI developed and used solely for scientific research. Standard commercial SaaS products do not qualify for these exemptions.
The Risk Class Determines Your Obligations
Not every obligation applies at the same time. The AI Act is structured around four risk categories: unacceptable risk (banned outright), high risk, limited transparency risk, and minimal risk.
If your product falls into the high-risk category defined in Annex III, the compliance burden is substantial. High-risk use cases include AI used in recruitment and CV screening, creditworthiness assessment, biometric identification, education and vocational training, and critical infrastructure management.
For high-risk systems, non-EU providers must meet requirements that include:
-
Conformity assessments before placing the system on the EU market
-
Technical documentation proving the system meets the requirements, including accuracy, robustness, and cybersecurity
-
Registration in the EU database for high-risk AI systems before placing the system on the market
-
Post-market monitoring and serious incident reporting
These high-risk obligations apply from 2 December 2027 for Annex III systems and 2 August 2028 for AI in products covered by Annex I, after the Digital Omnibus on AI (Regulation (EU) 2026/1744) moved them. SMEs and small mid-caps may provide the technical documentation in a simplified form (Article 11).
Limited-risk systems, such as chatbots, carry lighter obligations: primarily transparency requirements under Article 50, which have applied since 2 August 2026 and require people to be informed they are interacting with an AI. Generative systems must also mark their output as AI-generated in a machine-readable way; systems already on the market before 2 August 2026 have until 2 December 2026 for that.
Article 22: The Authorized Representative Requirement
This is where many non-EU companies get caught off guard. Article 22 requires that any provider established outside the EU who places a high-risk AI system on the EU market must appoint an authorized representative located in the EU before that system reaches European customers.
The authorized representative is not a formality. They carry real legal weight. Specifically:
-
They must be designated in writing by the non-EU provider.
-
They act as the single point of contact for EU market surveillance authorities.
-
They verify that the conformity assessment has been completed and that the required technical documentation exists.
-
They keep documentation available for the authorities and cooperate with them.
This is analogous to the GDPR's Article 27 representative requirement, which many US companies already handle through a European law firm or compliance service. The AI Act representative is a separate appointment and covers different obligations.
If you have a US-based SaaS platform used by 5,000 EU employees for performance appraisals or recruitment shortlisting, that system very likely qualifies as high-risk under Annex III. Without an authorized representative in place, you cannot legally place that system on the EU market once the high-risk provisions apply on 2 December 2027.
Key Dates Non-EU Companies Must Track
The AI Act entered into force on 1 August 2024. Obligations phase in over several years:
-
2 February 2025: Prohibited AI practices (Article 5) and AI literacy under Article 4 apply. If any feature of your product involves social scoring or workplace emotion recognition, this already applies. Article 4 requires providers and deployers to take measures to support the AI literacy of their staff.
-
2 August 2025: Obligations for General-Purpose AI (GPAI) model providers take effect.
-
2 August 2026: The Act applies in general, including the Article 50 transparency rules, and national market surveillance authorities supervise it.
-
2 December 2026: Two new prohibitions apply (AI that generates non-consensual sexual deepfakes or child sexual abuse material).
-
2 December 2027: High-risk obligations for Annex III systems apply: conformity assessments, EU database registration, and authorized representatives.
-
2 August 2028: High-risk obligations for AI in products covered by Annex I apply.
December 2027 feels distant. Conformity assessments, technical documentation, and governance processes take months to prepare properly, so start early.
What US SaaS Platforms Need to Do Now
If your product is used by EU customers, here is the practical sequence:
Step 1: Classify your system. Map every AI feature against the Annex III categories. This is not a legal opinion exercise; it is a product and compliance team exercise that should take two to four weeks.
Step 2: Check for prohibited features. Article 5 bans are already active. Subliminal manipulation, social scoring, workplace emotion recognition and certain biometric categorisation features are prohibited now.
Step 3: Appoint an authorized representative if required. If you have any Annex III use cases, you need a named entity in an EU member state. This should be a contracted service with clear accountability, not a loose arrangement with a local partner.
Step 4: Prepare technical documentation. Article 11 specifies what this documentation must contain, including system architecture, training data, performance metrics, and intended purpose. Start drafting now; it is not a short document.
Step 5: Build your conformity assessment process. Most Annex III systems follow the internal control procedure. Biometric systems (Annex III, point 1) may require a third-party notified body, depending on whether harmonised standards are applied.
Step 6: Register in the EU AI database. Before placing a high-risk system on the market, providers must register it in the EU database, which is largely public. Regulators and users will be able to look your product up.
Penalties Are Designed to Reach You
The fines under the AI Act are structured as percentages of global annual turnover. For violations involving prohibited practices, the maximum is €35 million or 7% of global annual turnover, whichever is higher. For high-risk system non-compliance, it is €15 million or 3%. For incorrect or misleading information provided to authorities, €7.5 million or 1%. For SMEs and start-ups, the lower of the two figures in each tier applies.
These figures apply to worldwide turnover. For a large company, the €15 million tier caps out at the higher of €15 million or 3% of worldwide turnover, even if its EU revenue is modest. EU market surveillance authorities can also require corrective action and restrict or withdraw non-compliant products from the EU market.
The Interaction With GDPR
Many non-EU companies already have GDPR compliance programmes in place. The AI Act works alongside GDPR, not instead of it. AI systems that process personal data must comply with both frameworks simultaneously. Where AI systems are used in high-risk contexts and involve personal data, your customers will often need a Data Protection Impact Assessment (DPIA) under the GDPR. Some of them, such as public bodies, providers of public services, lenders and life or health insurers, will also need a Fundamental Rights Impact Assessment (FRIA) under Article 27 of the AI Act from 2 December 2027, and they will ask you for the information to complete it.
If you already have an Article 27 GDPR representative in the EU, that person or entity cannot automatically serve as your Article 22 AI Act representative. These are separate legal appointments covering separate obligations.
One Action to Take This Week
Pull a list of every AI feature in your product that touches EU users. Mark each one against the Annex III category list. If anything appears on that list, schedule an authorized representative review before the end of the quarter. That single step gives you a clear plan for December 2027.
Not sure where your product sits on the risk spectrum? Run your AI system through the free 2-minute compliance check at comply.khairos.ai and get a clear starting point.
References
Not sure where your company stands?
Answer 15 questions about how your team uses AI. You get a compliance score and the gaps to close first, as a PDF.