The EU AI Act for small businesses: a short orientation
The EU AI Act for small businesses is mostly a matter of knowing where you stand. The Act is the European law on artificial intelligence. It sorts uses of AI by risk, and the duties grow with the risk. Most of what a small business does with AI, such as drafting text, sorting requests or summarising documents, falls into the lower levels, where the duties are light.
This is a general orientation, not legal advice. The rules take effect in stages and the details matter, so check your own case with an adviser.
The idea behind the law
The law does not judge the technology. It looks at what AI is used for. Think of a kitchen knife. In a kitchen it is a normal tool. In other places there are strict rules for carrying one.
It is the same with AI. The same AI can write a newsletter or sort job applicants. The law treats these two uses very differently, because the second one affects people's lives.
The risk levels, with small-business examples
| Level | Typical example | What it means for you |
|---|---|---|
| Not allowed | Uses seen as a threat to people's rights. Rare in normal business. | Do not do it |
| High risk | AI that helps decide about people, for example software that ranks job applicants | Many duties. Get advice before use. |
| Duty to be open | A chatbot on your website, images or voices made by AI | Tell people they are dealing with AI |
| Minimal risk | Drafting, summarising, sorting, search | No special duties |
Maker or user: which one are you?
The law gives different duties to different roles. Two matter here.
- A provider builds an AI system and offers it under its own name.
- A deployer uses an AI system in its own business.
A small company that uses the AI functions in its software is normally a deployer, in other words a user. The heavier duties sit with the providers.
This can change if you build an AI product and sell it. It can also change if you use an existing system for a high-risk purpose it was not made for. Then ask an adviser.
Four steps that make sense for any small company
- List where you use AI. Include the AI functions inside tools you already pay for, and what staff use on their own. Note for each entry what it is used for and whether it affects decisions about people.
- Sort the list by the table above. Most entries will be minimal risk. Mark anything that touches hiring, staff reviews, credit or similar decisions. Take those to an adviser.
- Be open where AI meets people. If customers chat with a bot, say so at the start. If you publish realistic images, audio or video made by AI, label them.
- Make sure staff understand the tools. The Act expects people who work with AI to know enough to use it sensibly. A short team session is a reasonable start. Cover what the tools can do, where they go wrong and what must not be typed into them.
Keeping a person in charge
For high-risk uses, human oversight is a formal duty. It is also a good habit everywhere else. Decide which AI results a person must approve. Keep a record of what the AI did. Name someone who can stop it.
What to ask your software suppliers
- Which AI does the product use, and where does it run?
- Can we see what the AI did and who approved it?
- Can we limit what the AI may see and change?
How Worknano handles this
Worknano gives you the parts of this that software can supply. You choose regional hosting in the EU, the USA or APAC, and workspace data and AI processing stay in that region. Assistants read only what the person using them may read. Changes they suggest are shown as previews, and nothing is saved until you confirm. Roles and access rules set the limits, and sensitive actions are recorded in a log.
Open a free workspace and keep your AI use where you can see and limit it.