AI literacy training and records
Who needs learning support, and why?
AI Act Article 4 concerns providers and deployers and their staff and other persons operating or using AI on their behalf. The AI literacy requirement has applied since 2 February 2025. As amended by Regulation (EU) 2026/1744 from 27 July 2026, it requires measures to support the development of AI literacy, taking account of those persons’ technical knowledge, experience, education and training, the context of use, and the persons or groups of persons on whom the AI systems are to be used. It does not prescribe one course for every organisation or guarantee a particular outcome for every employee.
Training is one practical way to support this requirement. Adapt the depth to each role: an everyday user, team leader, tool administrator and person responsible for consequential AI decisions have different needs.
Suggested content
- What the tool can do, where it fails and how to verify outputs.
- Approved tools and uses within the organisation.
- Protection of personal, confidential and contractually restricted information.
- Human oversight, escalation of errors and incident reporting.
- Relevant prohibited practices and transparency obligations.
- A task drawn from the participant’s work and an opportunity to ask questions.
Slide counts, question counts and pass marks depend on the material version and learning objective. They are not legal standards. The L1 — Align offer confirms the current scope of supplied materials.
Two delivery methods
| Method | Process | What to arrange |
|---|---|---|
| Facilitated training | Explanation, practical example, discussion and a check of understanding | A trainer, current organisational rules and participation by relevant people |
| Self-study | Materials, a practical task or quiz, and feedback | Accessible explanations, a contact for questions and follow-up learning |
An unsuccessful result should lead to explanation and practice of the difficult topic. Repeating identical questions alone is not reliable evidence of understanding.
What to record
A useful record identifies the participant and role, date, topic and material version, method of checking understanding, result and any follow-up learning. Check which information the selected tool actually stores and what the organisation must record itself. Do not assume automatic recording, export or centralised records without verifying the specific configuration.
Set access and retention according to the record’s purpose. Avoid collecting unnecessary personal data.
What an HMAC-protected code demonstrates
HMAC helps verify a record’s integrity and authenticity relative to the holder of a shared secret key. Reliability depends on implementation and key protection. It does not by itself establish the participant’s identity, actual attendance or a trustworthy timestamp. Another holder of the same key can generate a valid record.
Verification requires a defined record format, algorithm version, key management and verification procedure. Distinguish integrity checks from evidence of identity and participation. Independent verification without access to the original application requires a separately prepared mechanism.
When to update the content
Review content after a new AI use, a tool or policy change, a change in risk or lessons from incidents. An organisation can set a periodic review schedule; Article 4 does not prescribe a universal annual interval or a specific quiz.
Verified on 6 September 2026: AI Act Article 4, consolidated text, amending Regulation (EU) 2026/1744, RFC 2104: HMAC.