The Difference Between AI Awareness Training and Measurable AI Literacy
Most Lithuanian public sector institutions have now run some form of AI training. It covered what AI is, what the key risks are, and why staff should pay attention. Participants completed a quiz and received a certificate.
That is AI awareness training. It is not AI literacy, and it is not what Article 4 of the EU AI Act requires.
Key points
- AI awareness answers: "Do you know that AI has opportunities and risks?" AI literacy answers: "Can you work responsibly with AI in your specific role?" Article 4 requires an answer to the second question, not the first.
- Awareness training produces exposure to concepts. Measurable AI literacy produces evidence of specific competences, at specific proficiency levels, for specific roles and AI systems. Only the second satisfies an Article 4 compliance record.
- DigComp 3.0 converts "staff received AI training" into "staff demonstrated these specific competences at these levels." That is the difference between a completion certificate and a competence report.
- A simple test. Can your training provider state which DigComp 3.0 competences each module develops, at what proficiency level, before a learner starts?
What AI awareness training is
AI awareness training covers what AI is, how it works at a basic level, and what the key risks are. It is designed to give a broad population a shared baseline. Staff learn that AI systems can hallucinate, that they process data, that bias can affect outputs, and that legal frameworks govern their use.
This is useful. Institutions whose staff have no AI awareness at all are in a worse position than those who have run awareness sessions. The problem is not that awareness training is bad. The problem is confusing it with literacy.
Awareness training typically has four properties:
- Generic content applied to all staff regardless of role
- Delivered once, often as a mandatory e-learning module
- Assessed with a short quiz that tests recall of concepts
- Documented with a completion certificate recording a date and a name
None of these outputs tells you whether a procurement officer can evaluate an AI vendor's risk classification. None tells you whether a front-line case officer understands when to override an AI recommendation. None tells you whether a digital transformation lead can design a role-differentiated compliance programme. The certificate records attendance. It records nothing else.
What measurable AI literacy means
Measurable AI literacy is the ability to demonstrate specific competences at the proficiency level each role requires.
The distinction is precise. Awareness asks "Do you know that AI has opportunities and risks?" While Literacy asks "Can you work responsibly with AI in your role, recognise its limitations, assess its outputs, and know when to escalate?" (aiactblog.nl, May 2026)
Article 4 of the EU AI Act addresses the second question. The obligation is to take measures appropriate to each person's technical background, role, and the AI systems they use. "Knowing that AI exists" is not a contextual, role-differentiated competence. "Being able to identify when an AI-assisted decision tool has produced an anomalous output and document it correctly" is.
Measurable literacy has three properties that awareness training does not.
It is role-specific. A procurement officer evaluating AI vendors needs different competences than a front-line case officer using an AI tool daily. Measurable literacy defines which competences each role requires and assesses whether the learner has developed them.
It is assessed against a framework. DigComp 3.0 provides 500+ observable learning outcomes across 21 competences and 8 proficiency levels. A training programme mapped to these learning outcomes can show where a learner sits on the framework, not just whether they attended. (JRC144121, November 2025).
It produces documentation that names competences. A competence report names which DigComp 3.0 competences the learner demonstrated, at what proficiency level, and for what context. That is what an Article 4 compliance record needs to contain.
One scenario, two outcomes
Consider a front-line case officer at a Lithuanian social services agency. The officer uses an AI-assisted benefits eligibility tool that generates a recommendation for each application.
After an awareness programme, the officer knows that AI systems can produce biased outputs and that human oversight matters. They received a certificate and passed the quiz.
After a measurable literacy programme, the officer can demonstrate DigComp 3.0 Area 4 (Safety) at Intermediate Level 3 for their specific AI use context. They can identify anomalous output in the benefits system, document it using the institution's protocol, and escalate correctly. Their competence report names the specific competence, the assessment method, and the proficiency level.
When a regulator reviews the institution's Article 4 compliance record, the first officer's certificate says: "attended training." The second officer's competence report says: "demonstrated specific competences at a specific level, for this AI system, in this role context."
Those are different records. Only the second one answers Article 4's question.
The three layers of AI literacy measurement
A measurement approach that satisfies Article 4 needs to operate at three layers.
Layer 1: Learning outcomes. Did the specific knowledge and skill the training targeted transfer to the learner? Pre-training and post-training assessments against DigComp 3.0 learning outcomes provide this evidence. A quiz testing AI awareness concepts does not.
Layer 2: Application outcomes. Can the learner apply what they learned in their actual work context? This requires role-specific scenario-based assessment. For example, a procurement officer working through an AI vendor evaluation, or a case officer working through an AI output override scenario. Scenario-based assessment is more demanding to design than a quiz. But it is the only approach that demonstrates applied competence rather than recalled information.
Layer 3: Documentation outcomes. Is there a record linking the training to the AI systems the learner uses, the DigComp competences developed, and the proficiency level demonstrated? That record is what an enforcement inquiry will ask for.
Awareness training provides evidence only at Layer 1, and even there weakly. It provides evidence of recall, not competence. A programme that addresses all three layers produces documentation that can answer an enforcement inquiry.
What each approach produces as Article 4 evidence
The practical difference is clearest in what each approach leaves behind.
AI awareness training produces:
- A list of who completed the course and when
- A record of course content (what topics were covered)
- A pass/fail mark on a comprehension quiz
Measurable AI literacy produces:
- A needs assessment mapping each role to the AI systems it uses
- A record of which DigComp 3.0 competences each module addressed
- An assessment showing which competences each learner demonstrated, at what proficiency level
- A competence report linking training outcomes to the learner's role and AI use context
When a market surveillance authority asks whether your institution took proportionate, role-differentiated measures, the first list cannot answer. It records exposure. It does not record competence, proficiency level, or the specific AI use context the training addressed.
The second list can answer. It documents not just what training was delivered but why it was appropriate for those people, those AI systems, and those roles.
Four questions to ask before purchasing AI training
When evaluating training providers for AI literacy, four questions distinguish measurable literacy programmes from awareness training.
1. Which specific DigComp 3.0 competences does each module address? A provider who cannot answer this is delivering awareness training, not DigComp-aligned literacy training. The competence mapping should be available before purchase.
2. At what proficiency level is each competence addressed? A programme that addresses all competences at Foundation level does not serve staff operating at Intermediate level. The proficiency level must match the role's requirements.
3. What does the post-training assessment measure? A quiz testing whether staff can recall definitions is not an assessment of competence. An assessment of competence demonstrates that a learner can apply specific skills in context.
4. What documentation does the programme produce for each learner? The output must name specific DigComp competences, the proficiency level reached, and the assessment basis. A completion certificate with a date does not satisfy this.
Any provider who answers all four questions with specificity, and can show the competence mapping before purchase, is in the minority.
Why the gap matters more now
Article 4 enforcement started 2 August 2026. Until August 2026, an institution with only awareness training was in a grey zone. Most institutions were.
That grey zone has closed. National market surveillance authorities now have formal supervisory powers under the EU AI Act. The enforcement question is not "did you run training?" It is "did training produce role-appropriate literacy for the AI systems your staff use?"
Awareness training cannot answer that question. Measurable literacy programmes can.
This matters specifically for Lithuania's public sector, where DigComp 3.0 is named in the National AI Strategy 2026-2035 as the recommended measurement framework. Using DigComp 3.0 for compliance documentation is not a creative interpretation. The Digital Omnibus (Regulation (EU) 2026/1744, July 2026) explicitly references DigComp as a framework organisations can use for Article 4 compliance.
For institutions that have run awareness training and want to understand what they still need, the gap is addressable. Assess current competence against DigComp 3.0, identify gaps per role, fill those gaps with role-specific content, and produce a competence record that names what was demonstrated and at what level.
Kelias offers a free 6-month AI literacy platform for Lithuanian public sector, mapped to DigComp 3.0 and built for Article 4 compliance. Access at kelias.tech.
Ikpong Joseph Alexander holds an MSc in Artificial Intelligence and is the founder of Kelias. The primary regulatory source for this post is Regulation (EU) 2024/1689 Article 4, as amended by the Digital Omnibus (Regulation (EU) 2026/1744, July 2026). The awareness/literacy distinction draws on practitioner analysis at aiactblog.nl (May 2026). The primary source for DigComp 3.0 is JRC144121, available from the JRC publications repository.