What the law asks

What the law actually asks for.

The page to send your compliance lead or your board. Article 4 as it stands after July 2026, what the Commission says a programme should cover, and what is not required.

The River Liffey in Dublin with the dome of the Four Courts

Article 4 after the Omnibus.

The Digital Omnibus rewrote Article 4 in July 2026. The duty is now to take measures that support the development of AI literacy, and the text is explicit that you are not required to guarantee any particular level for any individual. What gets assessed is what your organisation did.

Two things follow. The European Commission states plainly that no certificate is required, and that organisations can keep an internal record of training and guidance. There is also no obligation to measure your employees' AI knowledge.

We build to the Commission's own published guidance. Its Q&A sets out four steps any Article 4 programme should cover. Here is how our workshops map to them.

The Commission's four steps for an Article 4 programme and where Karv Training workshops cover each.
aThe Commission asksWhat is AI, how does it work, what AI is used in our organisation, and what are its opportunities and dangers?Where we cover itOpening module: what AI is and how your tools use it
bThe Commission asksIs our organisation developing AI systems, or using systems built by others?Where we cover itRole and tool inventory
cThe Commission asksWhat do staff need to know about these systems, what risks, and what mitigations?Where we cover itRisk mapping and role-calibrated depth
dThe Commission asksBuild the literacy actions on that analysis, accounting for staff knowledge and the context of useWhere we cover itPolicy workshop and your written record

Who it applies to.

Article 4 binds every provider and deployer of AI systems, whatever the industry. It has applied since 2 February 2025, and national market surveillance authorities have supervised and enforced it since 2 August 2026. In Ireland that means your existing sectoral regulators, coordinated by the AI Office of Ireland.

  • 01

    Everyday AI tools

    Staff using ChatGPT or similar to write copy, summarise documents or translate. The Commission's own guidance confirms these organisations are in scope and must inform staff about specific risks such as hallucination.

  • 02

    AI inside software you already bought

    Recruitment screening, HR platforms, finance tools and customer service software increasingly ship AI features by default. You are a deployer whether or not you chose the feature.

  • 03

    Customer-facing AI

    Chatbots, deepfakes and certain AI-generated content carry separate transparency duties under Article 50, in force since 2 August 2026 and not deferred by the Omnibus.

  • 04

    Contractors and agencies

    The duty covers people who deal with AI on your behalf but are not on your payroll. Contractors, service providers and agencies should be covered by the same measures as your employees.

57.7%

of large Irish enterprises, with 250 or more staff, used AI in 2025. Across all enterprises with ten or more staff the figure was 20.2%. CSO, Information Society Statistics: Enterprises 2025

What about high-risk AI systems?

Article 4 applies now. The separate Article 26 duty on deployers of high-risk systems, such as AI that scores creditworthiness or screens job candidates, is to assign human oversight to people with the competence, training and authority to do it, and the Commission is explicit that relying on a system's instructions for use is not sufficient.

The Digital Omnibus moved most of those obligations to 2 December 2027, and a system already in service before that date is only brought in once its design changes significantly. Training for oversight roles is a separate, deeper engagement that we scope on a call.

A generic course cannot do step (d). Calibration to your tools, your roles and your context is the judgement Article 4 actually asks for, and the record of that judgement is what you keep.

This is general information about the EU AI Act, not legal advice. Enforcement is proportionate and case by case, and your sectoral regulator may issue guidance that applies to you.