Author and editorial responsibility
Tim Jamboula, Founder of Corporathon. Last reviewed 24 August 2026. This post is not legal advice.
Summary
Since 2 February 2025, Article 4 of the AI Regulation requires sufficient AI literacy for everyone who operates or uses AI systems, by role and context. AI training can build and document that competence but does not replace an official certification. What matters is a traceable record: who, which role, which content, which result. A hackathon delivers that record especially tangibly through a used artifact.
Table of contents
- The real question behind Article 4
- What Article 4 requires (and what not)
- The documentation frame in five fields
- Decision framework: which record for which role
- Honest cost logic of compliance
- A worked risk model (illustrative)
- Training, hackathon or both as evidence
- What to do next
- FAQ
1. The real question behind Article 4
"Do we satisfy the EU AI Act with an AI training" is the wrong expectation. Article 4 does not require a specific training but sufficient competence, appropriate to role and context. The right question is: can we evidence, for each relevant role, that the competence exists? Training is one path there, the record is the goal.
The EU AI Act does not ask for a certificate but for real, role-appropriate competence. It becomes evidenced through application, not attendance. – Tim Jamboula, Founder of Corporathon
2. What Article 4 requires (and what not)
| Aspect | What applies | Common misunderstanding |
|---|---|---|
| Applicability | since 2 February 2025, for providers and deployers | it applies now, not only later |
| Standard | sufficient competence, by role and context | one uniform certificate for all is not enough |
| Record | traceable documentation of measures | an attendance list alone evidences no competence |
| Responsibility | rests with the company | an external provider guarantees no compliance |
| Certificate | there is no mandated official document | a course slip is not an official record |
The core is adequacy. A specialist who operates an AI system productively needs more than someone who only uses it occasionally. The record must reflect that gradation.
3. The documentation frame in five fields
A robust record holds five fields per measure.
- Who. Names or roles of participants, tied to AI use.
- Which role and context. Deployer, user, department, risk class of the system.
- Which content. Topics, tools, limits and risks that were covered.
- Which result. For training, the knowledge; for a hackathon, the used artifact.
- When and by whom. Date, format, responsibles and, if external, the provider.
These five fields turn an attendance list into a traceable competence record.
Build your competence record on a real case
Two paths, depending on maturity level.
- Book directly: Book a discovery call. 30 minutes, we look at how you document competence by role.
- Read first: Enter your email and get the EU AI Act documentation template. No spam, unsubscribe any time.
4. Decision framework: which record for which role
Visualisation: required competence by role and system context (schematic, illustration without measured values).
required
competence
high | ● operator, high-risk system
|
| ● power user, productive use
medium|
| ● occasional user
low | ● viewer only
+------------------------------------------> Closeness to AI use
low high
Rule of thumb: the closer a role is to productive AI use, the deeper the record must reach. For deployers of higher-risk systems, an awareness talk is not enough, documented application counts here.
5. Honest cost logic of compliance
Compliance cost is rarely the course price, it is what surrounds it.
- Mapping. First you need to know who uses which AI systems in which role. That is effort before any training.
- Depth per role. Different roles need different measures, a one-size course is cheap but rarely appropriate.
- Documentation. Maintaining the record costs time, but saves far more in a review.
- Refresh. AI literacy is not a one-off state, refreshers belong in the calculation.
We quote prices personally. The right scope emerges in conversation, from these variables.
6. A worked risk model (illustrative)
Purely illustrative model, not a legal assessment. Assume a company makes a one-off investment in a role-appropriate competence program with clean documentation.
- Programme effort as an assumption: 25 internal hours of preparation plus the measure itself.
- At a 60 EUR internal hourly rate, the 25 hours are roughly 1,500 EUR of internal effort.
- Against this stands the risk of unclear competence: unproven records in a review, rushed rework under time pressure, possible sanction risk under the Regulation.
- The modelled comparison shows that early, documented effort usually sits well below the cost of late, hectic back-documentation.
The figures are an order-of-magnitude assumption, not a guarantee and not legal advice.
7. Training, hackathon or both as evidence
Training documents transferred knowledge and suits broad basic competence. A hackathon documents practical application through a used artifact and suits roles that use AI productively. For an appropriate overall program, the combination is often sensible: broad training as a base, a hackathon as deep evidence where productive work happens. Neither is an official certificate.
8. What to do next
First map who uses which AI systems in which role. Set the required depth per role and document along the five fields. For productive roles, a hackathon is the most tangible record because it creates application, one week from first contact to a used artifact. Assess the adequacy of the overall program yourself or with legal advice.
Next step
- Book directly: Book a discovery call. We look at your role and record needs.
- Read first: Enter your email and get the documentation template. Double opt-in, unsubscribe any time.
Related terms
EU AI Act · AI literacy · AI hackathon · AI training for companies
FAQ
Does AI training automatically satisfy the EU AI Act? No. Article 4 requires sufficient, role- and context-appropriate competence and a traceable record, not a specific certificate. Training can contribute; the company owns the adequacy of the overall program.
Since when does Article 4 apply? Since 2 February 2025. The duty of sufficient AI literacy affects providers and deployers of AI systems and applies already, not only in the future.
Is an attendance list enough as a record? Alone it evidences no competence. A record along five fields is sensible: who, which role and context, which content, which result, when and by whom.
Is a hackathon a competence record? It documents practical application especially tangibly through a used artifact. That is a strong building block, but not an official certificate and not automatic compliance.
Legal note: An AI training or hackathon can document AI literacy measures under Article 4, but are not an officially mandated certificate and do not automatically guarantee compliance with Article 4. This post is not a substitute for legal advice.