What is the EU AI Act and what must your business do now?

Does your team use ChatGPT, Copilot or an A.I. agent? Then the AI Act applies to you too. Not someday, but today. Many companies think the law is only about those who build A.I. That is not true: anyone who uses A.I. has obligations as well.

In this article you read what the AI Act is, which dates matter and what you need to have in place today. You get a 6-step plan and answers to the questions we hear most often.

What is the EU AI Act?

The EU AI Act, officially the Artificial Intelligence Regulation, is the European law on artificial intelligence. It applies directly in Belgium and the Netherlands, without any extra national law.

The AI Act is risk-based. The higher the risk of an A.I. application, the stricter the rules. A chatbot on your website follows different rules than an A.I. that screens job applicants.

Good to know: the law is not only for companies that build A.I. Do you use A.I. tools like ChatGPT, Copilot or your own A.I. agent? Then the law calls you a deployer. That is anyone who uses an A.I. system under their own responsibility. And deployers have obligations too. Only purely personal, non-professional use falls outside the law.

What are the 4 risk levels of the AI Act?

The AI Act sorts A.I. into four levels. Each level has its own rules.

• Unacceptable risk: A.I. that manipulates people or gives them a social score. Example: a system that scores citizens on their behaviour. Banned since 2 February 2025.
• High risk: A.I. that helps decide on people's chances. Example: a tool that screens job applicants or assesses the creditworthiness of individuals. Strict rules from 2 December 2027.
• Limited risk: A.I. that talks to people or creates content. Example: a chatbot on your website. You must say it is A.I.
• Minimal risk: most A.I. applications. Example: a spam filter or a tool that sorts your emails. No extra rules.

Most SMEs use A.I. with limited or minimal risk. The A.I. literacy duty does apply to everyone who uses A.I. professionally.

Which dates matter for your business?

Since / from	What does it mean for you?
2 February 2025	You take measures so your team learns to use A.I. responsibly (A.I. literacy). The banned A.I. practices also apply since this date. This applies now.
2 August 2026	A chatbot or A.I. agent that talks to customers must say it is an A.I. This applies now.
2 December 2026	New ban on A.I. that creates nude images without consent ("nudify" apps) or child sexual abuse material.
2 December 2027	Do you use A.I. for recruitment, staff evaluation or credit checks? Then strict extra rules apply.

So most companies need two things in place today: A.I. literacy and transparency.

What is A.I. literacy?

The AI Act gives its own definition. A.I. literacy is about the skills, knowledge and understanding to use A.I. systems in an informed way. Your team also needs to be aware of the opportunities, risks and possible harm that A.I. can bring.

In practice, it is about questions like:

• When can you trust an answer from ChatGPT, and when not?
• Which customer data can you put into an A.I. tool?
• How do you spot mistakes or bias in A.I. output?
• Who is responsible when an A.I. agent makes a mistake?

What exactly do you need to do for A.I. literacy?

The rule is in Article 4 of the AI Act. Since 27 July 2026, it is a bit more flexible, because of the Digital Omnibus (Regulation (EU) 2026/1744). This is what applies now:

• For whom: your own staff, but also freelancers or temporary workers who use A.I. on your behalf.
• What: you take measures so their A.I. knowledge grows. You do not have to guarantee a specific level, but you must be able to show your effort.
• Tailored: someone who handles customer data with A.I. every day needs more knowledge than someone who rewrites an email now and then.

In short: you do not need a perfect result, but you do need to show you are working on it. Doing nothing is not an option.

Who checks this in Belgium?

The BIPT will become the Belgian A.I. regulator. At the publication of this article, the law that arranges this is not approved yet.

Do not wait for it. The rules already apply. Do you have a complaint about A.I. today? Then the Belgian federal government advises you to contact the AI Act Service Desk of the European Commission, or to go to court (VRT NWS). Once the regulator is in place, you can go there too.

What if you use a chatbot or A.I. agent?

Since 2 August 2026, the transparency rules (Article 50 of the AI Act) apply. In practice:

• Chatbots and A.I. agents: people must know they are talking to an A.I., unless that is already obvious. This duty lies with the provider of the system. Note: do you have an A.I. agent built and use it under your own name? Then the law often sees you as the provider.
• Deepfakes: A.I. image, audio or video that resembles real people, places or events and could pass as real must be labelled as A.I. For clearly creative or satirical work, a light mention that does not spoil the work is enough.
• A.I. texts for the public: do you publish an A.I. text to inform people about matters of public interest? Then you disclose it. This is not needed if a person reviewed the text and someone holds editorial responsibility. Tip: always have your blogs and newsletters reviewed.
• Timing: give this information clearly, at the latest at the first interaction.

Do you use an A.I. agent that talks to customers? Then let it say so in its very first message. That way you are safe, because the law asks for it at the latest at the first interaction.

How do you make your company A.I. literate in 6 steps?

1. List your A.I. tools. Include tools people use without IT knowing.
2. Who uses what? Include freelancers and temporary workers.
3. Write an A.I. policy. Which tools, which data, who is responsible.
4. Train your team. Per role or department, with a workshop.
5. Document everything. Who followed which training and when. This is how you prove your effort.
6. Repeat every year. A.I. tools change fast, so plan an update.

How does Flowkify help?

Flowkify is a Belgian A.I. agency. We help companies in Belgium and the Netherlands use A.I. in a safe and smart way.

• A.I. workshops and training tailored to your team, per role or department
• A.I. governance: we help you with an A.I. policy and an overview of your tools
• A.I. agents and automations built with the AI Act in mind from day one

Book a free AI Act check. Together we look at which A.I. tools you use and what you still need to arrange.

Note: this article is not legal advice. Unsure about your situation? Ask a legal expert.