August 2026: The AI Act Obligations Irish Practices Cannot Defer

Obligation 1

Article 50(1) — Transparency at Every AI Touchpoint

From 2 August 2026, any AI system that interacts directly with a natural person must disclose its artificial nature at the earliest point of contact. The disclosure must be explicit and prominent.

For an Irish practice running an automated intake agent, a client-facing query handler, or a chatbot on its website, this obligation is structural. Embedding the disclosure in your terms and conditions does not satisfy the requirement. The Commission's draft guidelines published in May 2026 apply a circumspect consumer standard — the question is whether a reasonably observant individual would immediately understand they are interacting with an automated system, not a member of staff.

Audit every client-facing digital touchpoint now. If it involves AI interaction with a person, the disclosure obligation is live in ten weeks.

Obligation 2

Article 4 — AI Literacy

Article 4 requires providers and deployers of AI systems to take measures to ensure, to their best extent possible, a sufficient level of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf. This obligation entered into force on 2 February 2025. It is already active law.

The Omnibus has softened the standard — moving it closer to an obligation of means than a strict obligation of result. The obligation to actively support AI literacy across your workforce remains. The practical exposure is direct: a staff member who inputs unstructured client data into a public AI tool without understanding that input may be used for model training creates a live GDPR breach under Article 5(1)(b) on purpose limitation. Your practice's liability does not require a regulator to invoke the AI Act — it flows from GDPR independently.

The minimum defensible position is a written Internal AI Use Policy defining which tools are approved for use with client data, what data categories may be inputted, what is expressly prohibited, and what training staff have completed and documented.

Review your data processing agreements with all AI tool vendors. Confirm contractually that client data inputs are not used for model training and that all processing occurs within the EU.

Obligation 3

The Article 6(3) Self-Classification Exposure

Most AI tools used in Irish practices — document drafting assistants, practice management software with AI features, AI-assisted research tools — qualify for the Article 6(3) non-high-risk exemption on the basis that they serve a purely assistive capacity with a human operator in the loop.

The Omnibus has reinstated the obligation for providers to register AI systems in the EU database where they claim this exemption. More directly relevant to your practice: national market surveillance authorities retain the power to challenge a firm's self-assessment. If an inspector concludes that a tool crosses into high-risk territory — an AI tool parsing employee performance, screening client creditworthiness, or evaluating case outcomes — and you cannot produce contemporaneous documentation of how and when you assessed that classification, you face immediate escalation.

The required output is a Classification Memo for each AI tool in active use: what the tool does, why it falls under Article 6(3), what human oversight controls are in place, and when that assessment was made. Date it. File it.