Merytt Talk through your business
AI and regulation

What does the EU AI Act actually require of a small business?

If your firm uses AI tools rather than building them, the rules are shorter and calmer than the headlines suggest. Here is what applies, what does not, and one thing to do this week.

By Aidan SmaleWritten Six minutes to read

2 Feb 2025is when the AI literacy duty and the list of banned practices started to apply, according to the European Commission.
2 Aug 2026is when national authorities started enforcing AI literacy, the bans and the transparency rules, per the AI Act Service Desk timeline.
2 Dec 2027is the new date for the high-risk rules in Annex III, set by the Digital Omnibus on AI, Regulation (EU) 2026/1744.
No certificateis needed to show AI literacy. The Commission says an internal record of training or guidance is enough.

What does the EU AI Act actually require of a small business?

If you only use AI tools, you are a “deployer”. Your main duty is to help your staff understand the tools they use, including the risks, like made-up answers. You must not use banned practices. And some AI content shown to the public must be labelled. Most of the heavy rules are for the companies that build AI.

That is the short version. The rest of this piece goes through each part in plain words, so you can see where you stand.

Am I a provider or a deployer under the AI Act?

The Act sorts people by what they do with AI, not by how big they are. There are two roles that matter for most firms.

  • Provider: someone who develops an AI system, or has one developed, and puts it out under their own name or trademark. Paid or free.
  • Deployer: anyone using an AI system under their authority, unless it is for purely personal, non-professional use.

So a consultancy using ChatGPT, Copilot or an AI note taker is a deployer. Your staff using those tools for work counts. Someone using them at home for a holiday plan does not.

The line can move. If you pay a developer to build an AI tool and offer it under your own name, you may be the provider. Article 25 also says a deployer can be treated as the provider of a high-risk system if they put their name on it or change what it is used for. If that sounds like you, take advice.

What is the AI literacy requirement?

It is Article 4, and it applies to providers and deployers alike. It was rewritten by the Digital Omnibus on AI, which came into force in July 2026. The new text says you must:

“take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf”

You should take into account what people already know, and the context the tools are used in. The same article adds that this “does not require providers or deployers to guarantee any specific level of AI literacy of any individual.”

The Commission’s own Q&A makes it concrete. It asks whether a company whose staff use ChatGPT to write adverts or translate text must comply. The answer is yes, and that staff “should be informed about the specific risks, for example hallucination.”

“Other persons” matters too. The Commission says this can include contractors and service providers working on your behalf.

The duty is to help people understand the tools. It is not to turn them into experts.

Do I have to tell customers when they are talking to AI or seeing AI content?

Sometimes. The transparency rules sit in Article 50, and they have applied since 2 August 2026. They split the work between providers and deployers.

  • Chatbots: the provider must make sure people are told they are dealing with AI, unless it is obvious. If you bought the chatbot, that is mainly the vendor’s job. If it was built for you under your name, it may be yours.
  • Deepfakes: if you use AI to create or change images, audio or video that could pass as real, you must say so. Clearly creative or satirical work only needs a light-touch label.
  • Public interest text: if you publish AI-written text to inform the public on matters of public interest, you must disclose it. That does not apply if a person reviewed it and someone holds editorial responsibility.
  • Emotion recognition or biometric categorisation: if you use these, you must tell the people exposed to them.

Our view, not the law: if a customer is chatting to an AI on your website, just tell them. It costs one line and it builds trust.

What AI uses are banned under the AI Act?

Article 5 lists practices that are not allowed at all. Most small firms will never come near them, but it is worth knowing the list. Banned uses include AI that:

  • uses hidden or deceptive techniques to push people into harmful decisions
  • exploits people’s age, disability or financial situation to cause harm
  • scores people on their social behaviour in ways that lead to unfair treatment
  • infers how staff or students feel at work or in education, except for medical or safety reasons
  • builds facial recognition databases by scraping images from the internet or CCTV
  • sorts people by biometric data to guess things like their race, religion or sexual orientation

The Digital Omnibus adds new bans on AI that generates non-consensual sexual deepfakes and child sexual abuse material. Those apply from 2 December 2026.

What is a small business not required to do?

This is where a lot of the worry falls away. According to the Commission’s AI literacy Q&A, you do not need:

  • a certificate. An internal record of training or guidance is enough.
  • an AI officer or an AI governance board. No specific structure is required.
  • to test or measure what each person knows about AI.
  • one fixed training course. There is “no one size fit all”, in the Commission’s words.

The high-risk rules are also further away than many people think. These cover things like AI used to filter job applications or evaluate candidates. After the Digital Omnibus, they apply from 2 December 2027 for the Annex III list, and 2 August 2028 for AI built into regulated products.

One caution. The Commission notes that simply asking staff to read a tool’s instructions “might be ineffective” in many cases. A short, real conversation beats a PDF nobody opens.

How does the AI Act work in Ireland?

The AI Act is an EU regulation, so it applies directly in Ireland. Ireland’s Regulation of Artificial Intelligence Act 2026 was signed into law on 21 July 2026, according to the Department of Enterprise, Tourism and Employment.

It set up the AI Office of Ireland as the central point of contact. Day-to-day enforcement is spread across existing sector regulators. The Department says the Irish Act “does not add to the obligations placed on regulated entities by the EU Regulation.” It also says further legislation, including for the Digital Omnibus changes, will come in the autumn.

The AI Office of Ireland suggests a sensible first step for any business: list the AI tools you use, work out which risk category they fall into, and check you are not using any banned practice.

What can I do this week?

Make a one-page AI tool list. It should not take long, and it is the base for everything else. This is our suggestion, built on the steps in the Commission’s Q&A.

  1. List every AI tool in use. Ask the team. Include the ones built into tools you already pay for, like email, notes and design apps.
  2. Note who uses each one, and for what. Drafting proposals is different from screening CVs.
  3. Mark anything a customer sees. A website chatbot, AI images in a campaign, AI-written articles.
  4. Write the main risk next to each. Made-up facts, client data going where it should not, wrong tone.
  5. Spend thirty minutes with the team on it. Then write down the date, who came and what you covered. That is your record.

If anything on the list touches hiring, staff monitoring or credit decisions, flag it. Those are the areas where the bigger rules will land in 2027.

If you would like a second pair of eyes on how AI fits into your day-to-day work, we are happy to talk it through.

Where this comes from.

Every rule and date here is from the law itself or an official EU or Irish government page. We do not write what we cannot point at.

  1. Regulation (EU) 2024/1689, the Artificial Intelligence ActEUR-Lex. The definitions of provider and deployer (Article 3), banned practices (Article 5), value chain rules (Article 25), transparency duties (Article 50) and the high-risk list (Annex III).
  2. Regulation (EU) 2026/1744, the Digital Omnibus on AIEUR-Lex, 8 July 2026, in force 27 July 2026. The new Article 4 on AI literacy, the new bans and the new dates for high-risk rules.
  3. Timeline for the implementation of the EU AI ActEuropean Commission AI Act Service Desk. The application and enforcement dates, updated for the Digital Omnibus.
  4. AI literacy, questions and answersEuropean Commission, last updated 20 October 2026. What Article 4 means in practice, including the ChatGPT example and what is not required.
  5. AI Office of Ireland established under the AI Regulation Bill 2026Department of Enterprise, Tourism and Employment, 30 July 2026. The Irish Act, the AI Office and the distributed model of regulators.
  6. Understanding the EU AI ActAI Office of Ireland. The suggested first step for businesses.

Last checked 28 October 2026. This is not legal advice. The AI Act is still being amended and explained, so check the official sources, and if your use of AI touches hiring, staff, credit or anything you build and sell under your own name, take advice from a lawyer.

Questions

Does the EU AI Act apply to a small business that only uses tools like ChatGPT?

Yes. A business using AI tools for work is a deployer under the Act. The main duties are to support the AI literacy of staff, to avoid banned practices, and to label certain AI content shown to the public. Most of the heavier rules fall on the companies that build AI.

What is the AI literacy requirement in the AI Act?

Article 4 says providers and deployers must take measures to support the AI literacy of their staff and others using AI on their behalf. After the Digital Omnibus on AI, no specific level is required. The Commission says staff using tools like ChatGPT should be told about risks such as hallucination.

Do I need an AI certificate or an AI officer?

No. The European Commission says there is no need for a certificate, and no specific governance structure such as an AI officer or board is required. An internal record of the training or guidance you have given is enough.

When do the AI Act rules apply?

AI literacy and the banned practices have applied since 2 February 2025. Enforcement of those, and the transparency rules, started on 2 August 2026. After the Digital Omnibus, the high-risk rules in Annex III apply from 2 December 2027, and those for regulated products from 2 August 2028.

Who enforces the AI Act in Ireland?

Ireland uses a distributed model. The AI Office of Ireland, set up under the Regulation of Artificial Intelligence Act 2026, is the central coordinating body and single point of contact. Existing sector regulators supervise and enforce the rules in their own areas.

Not sure where AI fits in your business?

Book thirty minutes with a founder. We will look at the tools you use and how they fit your day-to-day work. No pitch, and the decisions stay yours.

Or take the six-question diagnostic