The EU AI Act for Small Businesses: What Applies to You, and What Is Just Panic

Since 2 August 2026, part of the EU AI Act applies to almost every business that uses AI with its customers. For most small businesses the changes are small. Here is what you need to do, and what is just panic.

The quick answer

If a customer chats with an AI on your website, they need to know it. If you publish an AI made photo or video that looks real, you need to say so. And you should help your team understand the AI tools you use. For a small business, the rest changes very little.

What changed on 2 August 2026

The AI Act has rules on different levels. The strictest ones are for what the law calls high risk AI, for example software that sorts job candidates or decides who gets a loan. Those rules only start in December 2027.

What did start in August are the transparency rules. The idea is simple: people have the right to know when they are dealing with AI (source: Cooley, August 2026).

The 3 things that matter for a small business

1. Your chatbot has to say it is AI

If customers chat with an AI on your website, WhatsApp or Messenger, they need to know they are not talking to a person, unless that is already obvious. One clear sentence at the start is enough.

Legally, this duty sits mainly with the company that provides the chatbot. But your customer doesn't see the difference. If it runs on your website, check that it says so. Most good tools have a setting for this.

2. Realistic AI images and videos need a label

If you use AI to create a photo, a video or a voice that looks or sounds real, you have to say it was made or changed with AI. Think of a "happy customer" who doesn't exist, or a person saying something they never said. A small "Made with AI" in the image or the caption does the job. Clearly artistic work, like an illustration, gets lighter treatment (source: Jones Walker).

3. Help your team understand AI

The law asks businesses that use AI to support AI knowledge among their staff. In a small team, that can be a short conversation and half a page that says which tools are allowed and what should never be pasted into them, like client data.

What is just panic

• Labelling every email or post you wrote with the help of AI. If you read it and you stand behind it, that is not needed.
• Registering somewhere or asking permission to use AI tools.
• Worrying about internal use: sorting invoices, writing drafts or summarising meetings is not touched by these transparency rules.

AI generated text on matters of public interest, a bit like news, needs a label. But not when a person reviewed it and takes responsibility for it. So your blog, your newsletter and the emails you read before sending are fine.

What about fines?

For these rules, fines can go up to €15 million or 3% of worldwide yearly turnover. That is a maximum, not a price. For SMEs the lower of the two amounts applies, and the law asks for fines to be proportionate.

Your 15 minute checklist

• List every place where AI talks to your customers: website chat, WhatsApp, automatic email replies, phone.
• Check that each one clearly says it is AI in the first message.
• Make sure a customer can always reach a real person.
• Look at your ads and social posts. Any realistic AI photos or videos? Add "Made with AI".
• Keep a human check on anything you publish in your company's name.
• Write half a page for your team: which AI tools you use and the basic rules.
• Using AI to screen CVs or make decisions about people? That falls under the strict rules from December 2027. Get advice before then.

How we handle this in our projects

When we build an AI assistant or an automation for a client, three things are in place from day one: the line that says it's AI, an easy way to reach a person, and a record of what the assistant said. Done at the start, it takes ten minutes. Fixing it later always takes longer.

This article is general information to help you understand the rules. It is not legal advice. For your specific situation, talk to a lawyer.