EU AI Act Explained: Risk Levels, Dates and Fines | Compleye

Last updated: November 2, 2026 

The EU AI Act (Regulation (EU) 2024/1689) is the world’s first comprehensive AI law, and it regulates AI systems according to their risk. Banned practices have applied since February 2025, general-purpose AI rules since August 2025 and most other rules since 2 August 2026, while high-risk obligations follow in December 2027 and August 2028. It applies to any company that offers or uses AI in the EU.

TL;DR

  • The AI Act uses a risk-based system with four levels: unacceptable (banned), high, limited and minimal risk.
  • High-risk AI, for example in hiring, education, credit scoring or healthcare, needs risk assessments, quality data, technical documentation, human oversight and strong cybersecurity.
  • Chatbots and generative AI fall under limited risk: users must be told they are dealing with AI, and AI-generated content must be labelled.
  • The Act applies to any organisation, inside or outside the EU, that places AI systems on the EU market or uses them in the EU.
  • Fines reach up to €35 million or 7% of global annual turnover for banned practices.

Not sure which AI Act risk level applies to you?

Compleye classifies your AI systems, builds a lean compliance roadmap and prepares the documentation the AI Act requires.

Book a meeting

Table of contents

The European Union has made waves with the AI Act (Regulation EU 2024/1689), the world’s first comprehensive AI regulation. As it rolls out, it’s reshaping the way AI is built, deployed, and regulated across the EU.

For businesses in AI, it’s not just about compliance, it’s about leading the charge. Aligning with the AI Act gives you a chance to build trust, show ethical responsibility, and stay ahead of the competition in a fast-changing digital world.

So, how should your AI-driven business navigate this game-changing law? Let’s break it down.

Understanding the EU AI Act: Four Risk Levels

The EU AI Act sorts AI systems into four risk levels: unacceptable risk (banned), high risk (strict requirements), limited risk (transparency duties) and minimal risk (no extra obligations).

Risk levelExamples from this articleWhat the AI Act requires
Unacceptable riskSocial scoring by public authorities, manipulative AI, untargeted scraping for facial recognition databases, emotion recognition at work or schoolBanned
High riskAI in autonomous vehicles, medical diagnosis, exam scoring, CV screening, credit scoring, law enforcement, border controlRisk assessment, quality data, technical documentation, human oversight, cybersecurity and robustness
Limited riskChatbots, AI-generated images, video and text, deepfakesTell users they are dealing with AI and label AI-generated content
Minimal or no riskSpam filters, AI in video games, recommendation algorithmsNo additional obligations

At the heart of the EU’s AI Act is a risk-based system that evaluates AI based on its potential harm. It strikes a balance between innovation and safety, ensuring AI systems are both advanced and trustworthy, while safeguarding fundamental rights. So, how are those risks determined?

Unacceptable Risk: Banned Practices

At the top of the AI risk scale are systems that are deemed an unacceptable risk, the ones that clearly threaten people’s safety, rights, or freedoms. These systems are simply banned under the AI Act, reflecting the EU’s strong commitment to protecting human dignity and keeping technology from being misused.

This includes AI that manipulates or deceives people to influence their behavior in harmful ways or targets vulnerable groups like children and individuals with disabilities. One of the most notable bans is on social scoring by public authorities, systems that assign a reputation score to individuals based on their behavior or characteristics, opening the door to unfair treatment or discrimination.

Other banned practices include predictive policing, where AI tools make criminal behavior predictions based on profiling, and the untargeted scraping of online content or surveillance footage to build facial recognition databases, a tactic seen as intrusive and privacy-violating.

Additionally, the use of emotion recognition in places like schools or workplaces is outlawed, as is biometric categorization that infers sensitive attributes such as race, political views, or religious beliefs. Finally, real-time biometric identification in public spaces by law enforcement is mostly prohibited, with only a few very strict exceptions allowed.

High Risk: Heavily Regulated AI

High-risk AI systems are those that could have a significant impact on people’s health, safety, or fundamental rights. While these systems aren’t banned, they come with stringent regulations before they can be used in the EU. Providers must perform thorough risk assessments, use unbiased and high-quality data, maintain detailed technical documentation, ensure human oversight, and implement strong cybersecurity and robustness measures.

These rules apply across various industries. In transportation, for example, AI used in autonomous vehicles or air traffic management must meet the highest safety standards, because a failure could literally cost lives. In healthcare, AI tools like robot-assisted surgery or diagnostic algorithms must be carefully evaluated to ensure they don’t cause harm or discrimination.

In education, AI systems that score exams or influence academic decisions must be fair, transparent, and reliable. In employment, tools used to screen resumes or assist with hiring must be designed to eliminate bias and ensure equal opportunity.

Access to essential services is also a high-risk area. For instance, credit-scoring systems that determine loan eligibility must be clear and non-discriminatory. In law enforcement, AI used to assess evidence or identify suspects can deeply affect individuals’ rights, so its use must be tightly regulated.

Similar safeguards apply in migration and border control, where AI handles visa or asylum requests, and in justice, where AI may support legal decision-making or analyze court rulings.

Limited Risk: Transparency Required

AI systems that interact directly with people but don’t pose significant risks fall into the “limited risk” category. While these tools don’t face the heavy regulations that high-risk systems do, they still need to meet transparency requirements to ensure users are fully informed. The aim is to maintain trust and support users in making informed decisions by clearly indicating when they’re engaging with AI.

For example, when users interact with chatbots, they must be informed that they’re chatting with an automated system, not a human. Similarly, generative AI tools, like those that create images, videos, or text, must label their content as machine-generated, especially important when dealing with deepfakes, which can convincingly mimic real people.

Transparency is even more crucial in public communication. If AI generates content like news articles or political messages, it must be clearly labeled as AI-produced. This ensures that the public can easily tell the difference between human and machine-generated content, fostering accountability in the digital space.

Minimal or No Risk: Largely Unregulated

Most AI systems currently in use fall under the minimal or no risk category, meaning they pose little to no threat to users’ rights or safety and therefore are not subject to additional obligations under the AI Act. These applications are typically used in everyday consumer technology and operate in low-stakes environments.

Think of AI-enhanced video games, where intelligent behavior is programmed to create a more immersive gaming experience, or spam filters that help keep inboxes clean by automatically sorting unwanted emails. Similarly, recommendation algorithms that suggest music, movies, or products based on user preferences are considered low risk. While these systems are widely used, they don’t require the same level of oversight as high-risk AI because their impact on individuals’ rights and freedoms are minimal.

Who Must Comply With the EU AI Act?

The AI Act has a broad territorial scope. It applies to any organization, whether based in the EU or not, that places AI systems on the EU market or uses them within the EU. This ensures that all AI affecting EU citizens is subject to the same standard of trust and accountability.

EU AI Act timeline: key dates after the Digital Omnibus

The EU AI Act applies in stages, and the Digital Omnibus on AI, Regulation (EU) 2026/1744, moved the high-risk deadlines to December 2027 and August 2028.

DateWhat applies
2 February 2025Bans on prohibited AI practices and AI literacy requirements
2 August 2025Obligations for providers of general-purpose AI models
2 August 2026General application date, including transparency duties such as telling users they are talking to AI
2 December 2026Marking of AI-generated content for systems already on the market before 2 August 2026, and safeguards against AI that generates non-consensual intimate content
2 December 2027High-risk obligations for stand-alone AI systems, for example in employment, education and credit scoring
2 August 2028High-risk obligations for AI embedded in regulated products such as medical devices

EU AI Act Compliance Challenges and Considerations

The AI Act provides a solid framework, but making sure your business is fully compliant, especially when it comes to high-risk AI systems, requires more than just checking boxes. It’s a journey, and there are a few bumps along the way.

First things first: Is your AI system high-risk? It’s not as simple as flipping a coin. You’ll need to dig into how the system’s used, which sector it operates in, and what impact it could have on people’s safety or rights. Think of it as matchmaking, but for AI and risk assessment.

Next up, documentation. High-risk AI systems don’t just show up with a user manual. They need a full dossier: how they were built, how they work, and how they meet the AI Act’s requirements. So, be prepared to hand over a detailed “life story” of your system when regulators come knocking.

Data, oh, data. It’s more than just numbers; it’s the backbone of your AI. Your training data has to be spot-on: accurate, relevant, and bias-free. After all, AI decisions can have serious consequences. Get the data wrong, and you risk not only compliance issues but also a public relations nightmare.

Then there’s human oversight. No, the machines aren’t ready to take over (yet). The AI Act insists that human operators stay in the loop, especially for high-risk systems. So, you’ll need to make sure there’s a way for people to monitor and intervene if things go awry.

Lastly, compliance doesn’t disappear once your AI is live. Post-launch, you’ll need to keep tabs on performance, spot any hiccups, and take action if something goes sideways. It’s like having a pit crew on standby for your AI just in case it needs a tune-up.

The Consequences of Falling Short: EU AI Act Fines

The EU AI Act has three fine levels: up to €35 million or 7% of worldwide annual turnover for banned AI practices, up to €15 million or 3% for breaching most other obligations, and up to €7.5 million or 1% for giving authorities incorrect or misleading information, with the lower amount applying to SMEs and start-ups.

Non-compliance carries serious consequences including fines of up to €35 million or 7% of global annual turnover, whichever is higher. Beyond financial penalties, businesses also risk damaging trust and credibility in an increasingly regulated and values-driven market.

Why Struggle with AI Compliance When You’ve Got Compleye?

The AI Act is here, and let’s be honest: it’s a lot. It’s packed with legalese, technical requirements, and more compliance hoops than you can shake a stick at. But don’t panic, Compleye is here to guide you through it all, making the complex world of AI regulations much more manageable, understandable, and even a little bit fun.

Step 1: Risk Assessment: Clarity from the Start

What’s your risk level? Low, high, or somewhere in between? Understanding where your AI stands isn’t always simple, but we’ve got you covered. Our team will dive deep into your AI’s functionality and give you a clear breakdown of its risk classification: no guesswork, just certainty.

Step 2: Tailored Compliance Strategy: Your Custom Roadmap

Compliance isn’t one-size-fits-all. Your AI, your industry, your goals, each is unique. We create a lean and practical compliance strategy that fits your business like a glove. Whether it’s new governance systems, workflow changes, or audit prep, we’ll guide you through it with a clear, actionable plan.

Step 3: Documentation Support: Paperwork, Made Simple

High-risk AI needs solid technical and legal documentation. We’ll handle the complex paperwork, showing exactly how your system works, how it was built, and how it meets the AI Act. No more hassle, just expertly prepared, regulation-ready documentation.

Step 4: Training & Workshops: Making Compliance a Team Effort

Compliance isn’t just for the legal team. We offer hands-on workshops to get your whole team up to speed on the AI Act. We break it down without the jargon, just clear, actionable insights to keep your entire organization compliant and on track.

Step 5: Ongoing Monitoring: Compliance Doesn’t End at Launch

The AI Act doesn’t stop at launch, it wants continuous monitoring. We’ll help you track your AI’s performance over time, ensuring it stays compliant through updates, changes, and audits. We’re here to keep you on track as your AI evolves.

Turning Compliance into a Competitive Edge

Let’s be honest: AI compliance is not just a checkbox, it’s a game-changer for your business. Aligning with the AI Act shows your commitment to data security, fairness, and transparency, setting you apart as an ethical leader in the AI world.

At Compleye, we help you turn compliance into a strategic advantage. By embracing ethical AI, you not only avoid penalties but also build trust, enhance your reputation, and secure your future. Stay ahead of the curve: it’s your new competitive edge.

This article is general information, not a substitute for advice from a qualified professional. Check the current requirements with your legal advisor before acting.

Frequently Asked Questions

What is the EU AI Act?

The EU AI Act, Regulation (EU) 2024/1689, is the world’s first comprehensive law on artificial intelligence. It regulates AI systems according to their risk to health, safety and fundamental rights: some uses are banned, high-risk systems must meet strict requirements, and chatbots and generative AI must be transparent. It applies to any organisation that places AI on the EU market or uses it in the EU.

The EU AI Act applies in stages. Bans on prohibited practices have applied since 2 February 2025, general-purpose AI rules since 2 August 2025 and most other rules since 2 August 2026. After the Digital Omnibus on AI, high-risk obligations apply from 2 December 2027 for stand-alone systems and from 2 August 2028 for AI embedded in regulated products.

Yes. The EU AI Act applies to any organisation, wherever it is based, that places AI systems on the EU market or puts them into service in the EU, and to providers and deployers outside the EU whose AI output is used in the EU. A US or UK company selling an AI-powered product to EU customers must comply in the same way as an EU company.

EU AI Act fines go up to €35 million or 7% of worldwide annual turnover for banned AI practices, up to €15 million or 3% for breaching most other obligations, and up to €7.5 million or 1% for giving authorities incorrect or misleading information. For large companies the higher amount applies. For SMEs and start-ups, the lower amount applies.

Usually not. Most chatbots fall under limited risk, which means you must tell users they are talking to an AI system and label AI-generated content. A chatbot becomes high-risk only if it is used for a high-risk purpose listed in the Act, for example screening job applicants, scoring exams or deciding on access to credit or essential public services.