The EU AI Act
for small businesses
Most small businesses are deployers: they use AI that someone else built. Their main duties are staff AI literacy and, from 2 August 2026, telling people when they deal with an AI. High-risk rules, such as for CV screening, now apply from 2 December 2027 after the 2026 Digital Omnibus amendment.
Are you a provider or a deployer?
The EU AI Act (Regulation (EU) 2024/1689) gives different duties to different roles. The two roles that matter for most businesses are:
- Deployer: an organisation that uses an AI system under its own authority. A business that uses ChatGPT for drafts, or subscribes to an AI receptionist product, is a deployer.
- Provider: an organisation that develops an AI system, or has it developed, and places it on the market or puts it into service under its own name or trademark. A software company that sells an AI product is a provider.
The line is less obvious when you have a custom agent built for your own use. If the agent is developed for you and you put it into service under your name, you can be the provider of that system, even if a studio wrote the code and the model comes from OpenAI or Anthropic. Agree in writing who carries which duty before the build starts. Our vendor questionnaire includes these questions.
The four risk levels, with small business examples
The Act sorts AI uses by risk. Most small business uses are in the lower two levels. The level depends on what the system is used for, not on which model it uses.
| Level | SME examples | What it means for you |
|---|---|---|
| Prohibited (Article 5) | Emotion recognition of staff at work, social scoring, manipulative techniques that cause harm | Do not use. These bans have applied since 2 February 2025. |
| High-risk (Annex III) | CV screening or ranking of job candidates, credit scoring of individuals, AI that decides access to essential services | Strict duties for providers and deployers from 2 December 2027. |
| Transparency (Article 50) | Website chatbot, WhatsApp agent, AI receptionist, AI-generated images of real people | Tell people they deal with an AI, and label deepfakes. From 2 August 2026. |
| Minimal risk | Spam filters, internal drafting help, stock forecasts, summarising documents | No specific AI Act duties beyond AI literacy. |
An AI receptionist or customer service agent is normally in the transparency level. A recruitment tool that filters or ranks applicants is high-risk, because Annex III lists AI used for recruitment and selection. That includes tools bought as part of an applicant tracking system.
Which dates apply, after the Digital Omnibus?
In July 2026 the EU adopted Regulation (EU) 2026/1744, often called the Digital Omnibus on AI. It was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. It moved the high-risk dates and changed several other rules. It did not move the Article 50 transparency date.
| Date | What applies |
|---|---|
| 1 August 2024 | The AI Act enters into force. |
| 2 February 2025 | Prohibited practices (Article 5) and the AI literacy duty (Article 4). |
| 2 August 2025 | Rules for general-purpose AI model providers, governance and penalties. |
| 27 July 2026 | The Digital Omnibus enters into force. The softer Article 4 AI literacy wording applies. |
| 2 August 2026 | Article 50 transparency duties. National market surveillance authorities supervise and enforce. |
| 2 December 2026 | New bans on AI that creates non-consensual intimate images or child sexual abuse material. Generative AI systems on the market before 2 August 2026 must meet the Article 50(2) marking duty. |
| 2 August 2027 | Each Member State must have at least one national AI regulatory sandbox in operation. |
| 2 December 2027 | High-risk duties for Annex III systems, such as recruitment, credit scoring and education. |
| 2 August 2028 | High-risk duties for AI in products covered by Annex I, such as machinery and medical devices. |
What a deployer must do
If you only use AI, your duties are short. They still need an owner in the business.
- AI literacy (Article 4). Take measures to support the AI literacy of staff who use AI for you. The Omnibus removed the “sufficient level” wording, but the duty remains. See our AI literacy guide.
- Transparency (Article 50). From 2 August 2026, people must know when they talk to an AI, unless it is obvious. If you use emotion recognition or biometric categorisation, you must tell the people exposed to it. You must label deepfakes. See our disclosure guide.
- Do not use prohibited practices. Check that no tool you buy does emotion recognition of staff in the workplace.
- High-risk use, from 2 December 2027. Use the system as the provider instructs. Assign human oversight to trained staff. Keep the logs the system makes for at least six months. Tell workers and their representatives before you use a high-risk system at work. Tell people when a high-risk system helps make decisions about them.
GDPR applies in parallel to all of this. The AI Act does not replace your privacy notice, lawful basis or DPIA. Our GDPR guide for AI agents covers that side.
What relief does the Act give small businesses?
The Act has specific measures for small and medium-sized enterprises (SMEs), and the Digital Omnibus extended some of them to a new group of small mid-caps (SMCs): companies with fewer than 750 employees that are not SMEs.
- Lower fines. For SMEs, each fine is capped at the lower of the fixed amount and the turnover percentage. For other companies it is the higher. The Omnibus applies the same rule to SMCs.
- Simpler documentation. SME and SMC providers of high-risk systems can provide technical documentation in a simplified form.
- Proportionate quality management. Quality management system duties are scaled to the size of the organisation.
- Priority sandbox access. SMEs, including start-ups, get priority access to AI regulatory sandboxes.
| Infringement | Maximum fine |
|---|---|
| Prohibited practices (Article 5) | €35 million or 7% of worldwide annual turnover |
| Most other duties, including Article 50 transparency and deployer duties | €15 million or 3% of worldwide annual turnover |
| Incorrect or misleading information to authorities | €7.5 million or 1% of worldwide annual turnover |
In Ireland, fines are imposed through the adjudication process in the Regulation of Artificial Intelligence Act 2026, and the High Court must confirm them.
What to do this quarter
- List every AI tool and agent the business uses, who owns it and what it is used for.
- Mark your role for each one: provider or deployer.
- Mark the risk level for each use. Flag anything in recruitment, credit or access to services.
- Check every customer-facing agent for a clear AI disclosure at the start.
- Run and record basic AI literacy training for staff who use AI.
- For any high-risk use, plan the work now. 2 December 2027 is close for a procurement cycle.
If you plan a new agent, build these decisions into the spec. We do this as standard on AI agent projects, and we will tell you if a planned use moves you into the high-risk level.
Common questions
Does the AI Act apply to a small business that only uses ChatGPT?
Yes, but lightly. A business that uses ChatGPT is a deployer. You must take measures to support AI literacy among staff who use it, and you must not use AI for prohibited practices. If you publish AI-generated deepfakes, or let customers talk to an AI without knowing, the Article 50 transparency duties also apply.
Is an AI receptionist high-risk under the AI Act?
Normally no. An AI receptionist that answers calls, books appointments and answers questions is in the transparency level. From 2 August 2026 callers must be told they are talking to an AI, unless it is obvious. It would move up only if it made decisions in an Annex III area, such as access to essential services.
When do the high-risk rules apply now?
After Regulation (EU) 2026/1744, high-risk duties for Annex III systems, such as recruitment and credit scoring, apply from 2 December 2027. High-risk AI in products covered by Annex I, such as machinery, follows from 2 August 2028. The original date for both groups was 2 August 2026.
What are the fines for a small business?
The maximums are €35 million or 7% of turnover for prohibited practices, €15 million or 3% for most other duties, and €7.5 million or 1% for misleading information. For SMEs the lower of the two figures applies. National authorities must also consider the economic viability of SMEs when they set penalties.
Sources
- EUR-Lex: Regulation (EU) 2024/1689 (AI Act)
- EUR-Lex: Regulation (EU) 2026/1744 (Digital Omnibus on AI)
- European Commission: AI Literacy Q&A
- European Commission: Transparency obligations under Article 50 FAQ
- Oireachtas: Regulation of Artificial Intelligence Act 2026 (No. 31 of 2026)
We last checked these sources on 19 November 2026. Rules, fees and prices change. Check the source before you act.
Bring us the task.
Tell us what happens today, which tools are involved and where it goes wrong. On a free 30-minute review, we’ll work out whether an agent, an automation or neither is the right fix, and what a sensible first release would be.
Start a project.

Book a call or send us a message. We read everything.
tash@neu.ie
Working with teams in Ireland and worldwide.
Free 30-minute AI review.