The AI Act after the Omnibus: what a 30-person Irish firm actually has to do

If you run a 30-person firm that uses ChatGPT or Copilot, the EU AI Act asks less of you right now than the headlines suggest. But it doesn't ask nothing, and the parts that apply are worth doing anyway.

What changed in 2026

The EU's Digital Omnibus on AI came into force on 27 July 2026. Its biggest effect was to postpone the rules for high-risk AI. Uses such as recruitment, credit scoring and education now face those obligations from 2 December 2027, and AI built into regulated products such as medical devices from 2 August 2028.

It also softened the AI-literacy duty in Article 4. The original text required organisations to ensure a "sufficient level" of AI literacy among staff. The amended version requires them to take measures to support the development of their staff's AI literacy. The duty itself still applies, and has since February 2025.

What applies to a typical SME today

AI literacy. If your staff use AI at work, you should be taking steps to help them use it competently, and be able to show it. Training records are the practical evidence.
Prohibited practices. A short list of uses, such as manipulative techniques and social scoring, has been banned since February 2025. Few SMEs come near these, but it's worth a two-minute check.
Transparency. Since 2 August 2026, if you use a chatbot with customers or publish AI-generated content, people should be told they're dealing with AI.

If you use AI to screen job applicants, assess creditworthiness or evaluate students, you're heading into high-risk territory. You have until December 2027, but preparing takes time.

In Ireland

The Government's plan, published in February 2026, gives Ireland a distributed model: a new AI Office of Ireland coordinating, with existing regulators such as the Central Bank, the HPRA and the CCPC supervising AI in their own sectors.

Four things to do this quarter

Find out what's in use. A short anonymous staff survey will usually turn up more AI tools, and more personal accounts, than you expect.
Write a one-page AI-use policy. Which tools are approved, what data may go into them, and when a person must check the output.
Train your people, and keep a record. Practical training on real tasks does the most good, and the attendance record covers your literacy duty.
Flag anything near the high-risk list. If AI touches hiring, lending or assessment decisions, start planning human oversight now.

None of this is about fear. Firms with a formal AI policy are ten times more likely to report major productivity gains from AI. The rules and the results point the same way.

Sources

General information, not legal advice.