Basil C. Puglisi

Artificial Intelligence (AI) & Digital Marketing Since 2009

Ireland Builds a National AI Literacy Taskforce While EU Law Sets No Passing Grade #AIg

by Leave a Comment

Officials seated at a signing table in front of a Department of Government Enablement Abu Dhabi and MBZUAI memorandum of understanding backdrop
Image: A memorandum of understanding signing between the Department of Government Enablement, Abu Dhabi, and MBZUAI, published with the AiNative training programme announcement, October 7, 2026. Source: Abu Dhabi Media Office.

Ireland is already testing generative AI that reads social welfare claim documents and helps staff make decisions. The national framework that will say what those staff, and everyone else, need to know about AI is due before the end of 2027. That gap is the whole AI literacy problem in one country’s plan, and it isn’t unique to Ireland.

The Intelligence:Applied programme, launched by the Department of the Taoiseach and published October 13, sets out 35 initiatives running to 2030. Among them: a €150 million AI Future Skills programme, a new National AI Literacy Taskforce under the AI Office of Ireland, and an AI Skills and Literacy Programme for schools that will support more than 43,000 primary teachers and 17,000 post-primary students. The same plan has a new Social Protection Innovation Hub piloting generative AI that processes claim documentation to “assist decision-making”, tested with staff in a sandbox, with public availability in 2027 “subject to testing/sign off”.

This column holds a firm line on what literacy should mean once AI touches a decision. It’s the competence to look at an AI output with real consequences and accept it, modify it, or reject it, then own that call by name. A training record can’t show that. When the people reviewing AI can’t catch its errors, review turns into Responsible AI in the narrow sense, machine checking machine with no real human review, and a certificate sits in the file beside it. AI Governance needs a person at the checkpoint who can say no and be right.

Ireland funds the skills push and hands the bar to a taskforce

The money and the reach are serious. AIReady.ie, the government’s free online skilling platform, moves to a next phase focused on AI as a role enhancement tool. A National Skills Academy for Digital Transformation and AI will coordinate employer demand. The Institute of Public Administration’s Digital Academy will offer AI learning pathways for the public service in 2027, building on what the release calls more than 20,000 AI training sessions already undertaken across the public sector.

Notice the unit. Sessions measure attendance. The Department of Enterprise, Tourism and Employment’s summit release says the taskforce will bring together industry, education, and public sector stakeholders to drive a coordinated national approach so people have “the skills and confidence” to engage with AI. Confidence is a fine goal for a citizen. For a claims officer working beside a document-processing model, confidence without the ability to spot a bad extraction is a risk, not a skill.

EU law now asks for effort, not a passing grade

Ireland’s plan sits inside the EU AI Act, and the floor under it just moved. The current text of Article 4, as amended by the Digital Omnibus on AI, still requires providers and deployers to take measures to support the AI literacy of their staff and others operating AI on their behalf. It now adds that the obligation “does not require providers or deployers to guarantee any specific level of AI literacy of any individual.” The Commission’s AI skills and literacy page puts it plainly: no specific, or “sufficient,” level is mandated, and national market surveillance authorities began supervising and enforcing the rule in August 2026.

The Commission’s AI literacy Q&A goes further. Article 4 “does not entail an obligation to measure the knowledge of AI of employees.” There’s “no need for a certificate.” Organisations can keep an internal record of trainings. The AI Board still has to adopt recommendations with common objectives under Article 4(3), and the Commission says that work is under way. Until it lands, a compliant literacy program can be a log of who attended what.

There’s one place the bar holds. Article 26(2) requires deployers of high-risk systems to assign human oversight to people “who have the necessary competence, training and authority, as well as the necessary support.” The Commission’s Q&A adds that relying on the instructions for use isn’t enough. And Annex III lists AI used by public authorities to evaluate eligibility for public assistance benefits as high-risk. Whether Ireland’s claims assistant meets that description is a classification call for the department. The lesson travels either way: the law asks for competence and authority exactly where AI shapes a decision about a person.

Two launches show the split between testing people and training on tools

The Wyoming Department of Education announced on October 15 a free AI Ready Skills Credential Program for schools with Prodigy Learning and Microsoft. Students work through four units, including Practicing Responsible AI, and show what they learned through formative and summative assessments that end in a capstone credential exam. Students learn inside Minecraft Education worlds, and the program aligns with state computer science standards for grades 6 to 8 and 9 to 12. Whatever you think of Minecraft as a classroom, the design measures something.

Abu Dhabi’s AiNative programme, announced October 7, takes the other route. Built with OpenAI, Microsoft, and MBZUAI, it aims to upskill 20,000 government employees by the end of the year on practical tools including ChatGPT, Codex, and Microsoft 365 Copilot, with function-specific certifications to follow in later phases. It follows a Copilot rollout to 35,000 government employees announced in July. Tool fluency matters. But a course built around a product teaches people how to use it, and that’s not the same as teaching them when to overrule it.

Put the passing grade where the decision happens

Basil’s HEQ/AIS work measures what he calls governance competence: whether a person can direct, challenge, verify, and own work done with AI. That definition is testable at the point of use, which is exactly where Article 4 has stopped asking for a level. He’s also argued that the EU and OECD AILit framework files the checkpoint inside learner self-monitoring, and the Omnibus recital names that same AI Literacy Framework for Primary and Secondary Education as one the Commission and Member States could take into account. If the common objectives inherit that design, literacy will describe awareness and leave verification optional.

Ireland still has room to do better than the floor. A national framework that defines literacy by role, and tests the roles that sign off on AI output, would give the taskforce something attendance counts can’t: evidence that the person at the checkpoint can catch the failure the system actually produces.

Three moves for anyone writing a literacy budget this quarter

Set a passing grade for every role that approves AI output. List the three to five failure modes that role has to catch, such as a misread field, an invented citation, or a wrong eligibility rule, and build a short test from real outputs with known errors seeded in. The measure is the seeded-error catch rate per reviewer, rerun each quarter. Pick the threshold before the test and hold it; anyone below it gets retrained before they sign off again.

Log decisions, not just completions. Record whether each reviewer accepted, modified, or rejected the AI output, and look at the pattern. A reviewer who accepts every output for a full quarter isn’t a success story. That’s a flag for passive acceptance, and it should trigger a spot audit of that reviewer’s queue.

Make refusal part of the curriculum, especially when a vendor builds it. Before anyone gets access to an AI tool tied to decisions about people, they should name the escalation path and practice at least one rejection on a realistic case. Track it as the share of users with a logged rejection drill before access, and the target is all of them. If you deploy in the EU, keep the Article 4 training record, then attach the test results to it.

Ireland’s claims pilot will face a sign-off before it reaches the public. The person who signs should be able to show they can catch what the model gets wrong, not just that they finished the course. The taskforce has until the end of 2027 to define literacy for a whole country. Any team deploying AI this quarter can define it for one decision by Monday.

Sources

FAQ

What is Ireland’s National AI Literacy Taskforce?

It’s a new taskforce established by the AI Office of Ireland under the Intelligence:Applied programme announced October 13, 2026. It will develop and deliver a national framework to embed AI literacy at all levels of society before the end of 2027, bringing together industry, education, and public sector stakeholders.

What does Ireland’s Intelligence:Applied programme include for AI skills?

It includes a €150 million AI Future Skills programme, the next phase of the free AIReady.ie platform, a National Skills Academy for Digital Transformation and AI, and an AI Skills and Literacy Programme for schools supporting more than 43,000 primary teachers and 17,000 post-primary students, with a 2030 goal of age-appropriate AI literacy for every student.

Does EU AI Act Article 4 require a specific level of AI literacy?

No. After the Digital Omnibus on AI, Article 4 still requires providers and deployers to take measures to support AI literacy, but it states the obligation does not require them to guarantee any specific level of AI literacy of any individual. The Commission says no certificate is needed and an internal record of trainings can document compliance.

Who enforces EU AI Act AI literacy rules?

National market surveillance authorities in each Member State supervise and enforce Article 4, starting in August 2026. The AI Board is due to adopt recommendations with common objectives under Article 4(3), and the Commission says that work is under way.

Where does the AI Act still require competence from people overseeing AI?

Article 26(2) requires deployers of high-risk AI systems to assign human oversight to people with the necessary competence, training, and authority, plus the necessary support. Annex III lists AI used by public authorities to evaluate eligibility for public assistance benefits among high-risk uses.

How can an organization measure AI literacy instead of counting training?

Set a passing grade for each role that approves AI output, test it with real outputs that contain seeded errors, and track the catch rate each quarter. Log whether reviewers accept, modify, or reject AI output, and require a practiced rejection drill before anyone gets access to AI tied to decisions about people.

AI disclosure: This column was researched and drafted with AI assistance by HAIA Agents (RAI by GrokBot), using primary sources verified at time of publication. #AIgenerated

Subscribe to Blog via Email

Enter your email address to subscribe

Join 9,585 other subscribers

Reader Interactions

Leave a Reply