AI training you can prove.
Your teams already use artificial intelligence. Since 2 February 2025, Article 4 of Regulation (EU) 2024/1689 has asked organisations to take measures to support the AI literacy of their staff, proportionate to each role, and it is advisable to be able to document them. CorpusProof's training delivers both: training by role and the record of the measures.
The training is delivered in Spanish, for firms in Spain and Latin America: the sessions, the materials, the assessments and the record are all in Spanish. This page describes it in English; the report and the samples it links to are in Spanish.
A one-size-fits-all course pays no attention to role or context
Article 4 asks for measures that take into account what each person knows and the context in which they use AI. An identical course for the whole workforce pays no attention to role or context. That is why we write each programme from scratch.
| The course you find on Google | CorpusProof's training | |
|---|---|---|
| What you buy | A pre-recorded syllabus, identical for everyone | Content written for your roles and your cases |
| Where the examples come from | From a manual | From your case files, your tools, your policies |
| Who it trains | The whole workforce, all the same | Each role gets its own, according to what it decides and the risk it takes on |
| What it assesses | That you know how to use the tool | That you know when not to use it, in the tasks of your role |
| What it leaves behind | Proof of attendance | A record of who was trained, when, on which version and with what result |
| If the tool changes | The course expires | The learning path carries on: the risk of the role hasn't changed |
Why we tailor it: our decision, supported by Article 4
This is not a future obligation. It has applied since 2 February 2025. Article 4 of Regulation (EU) 2024/1689 asks for measures that take into account "their technical knowledge, experience, education and training and the context the AI systems are to be used in". A syllabus recorded once for the whole workforce takes none of those four things into account: not the role, not prior experience, not the specific tasks in which each person uses AI.
The article also makes clear what it does not require: it does not set a specific level of AI literacy for any particular person. We say so just as clearly, because almost no one in the sector has caught up with the July 2026 amendment and many still sell on the old wording. That the rule sets no such threshold does not change our view: if it asks for each person's context to be taken into account, identical training for the whole workforce does not do that, and that is why we write every programme from scratch. This is our decision, not a reading the law imposes.
Don't take our word for it: here is the material (in Spanish) → For example, the sample for law firms (in Spanish).
What the rule says and what has already happened
- 2 February 2025The date from which Article 4 of Regulation (EU) 2024/1689 applies. The Digital Omnibus did not postpone it.
- No size thresholdIt applies to every organisation that uses AI systems, whatever its headcount or turnover.
- 4 rulings against professionalsIssued by Spanish courts between 2024 and 2026 for using AI without human oversight.
Each figure, with its source cited in the report (in Spanish).
The courts no longer warn: they sanction
Between September 2024 and July 2026, Spanish courts decided four cases against professionals who filed submissions with case law invented by generative AI tools. The Constitutional Court unanimously issued a formal warning to a lawyer who attributed 19 non-existent citations to the court itself, and referred the matter to the bar association (Press Release 90/2024, in Spanish). In February 2026 the High Court of Justice of the Canary Islands imposed a €420 fine for 48 invented citations (General Council of the Judiciary note, 13 February 2026, in Spanish) and, in July 2026, another of €840 on a lawyer who did not acknowledge the error.
None of those sanctioned acted in bad faith: they trusted the output of a general-purpose tool without checking it. The Canary Islands court described that omission as a "breach of the basic duty of human oversight". It is the kind of risk for which Article 4 of Regulation (EU) 2024/1689 has required, since 2 February 2025, taking AI literacy measures that take into account the role and the context of use. The law asks for those measures to be taken; documenting them is how you can show that they were.
Training built to stand as evidence
Role diagnostic
A map of the real use of AI in your organisation: who uses which tool, for which task and with what level of risk. It is included in the programme and sizes everything else; it starts from the free 20-minute questionnaire.
Learning paths by role
Training specific to each role, not a generic syllabus: built on the cases, the tools and the policies of your firm.
Assessment proportionate to risk
Whoever uses AI in tasks with effects on people is assessed more demandingly than whoever drafts internal documents. It is a proportionate way of taking into account the context of use that the rule asks for.
Record of the measures taken
Syllabus, attendance register, assessment results and versions of each piece of content: the document you show when someone asks.
You are asked what AI training your team has
The question doesn't come in the abstract: it comes from a client during an audit, from an insurer or from an authority. What changes is not the answer, it is what you can put on the table.
With a generic course
- An identical syllabus for the whole workforce.
- Proof of attendance.
- No record of what each role knows or which tool it works with.
Article 4 asks for each person's knowledge, experience and context to be taken into account; an identical syllabus for everyone leaves no record of any of that.
With a CorpusProof programme
- A learning path by role, built on your cases, your tools and your policies.
- Assessment proportionate to the risk of each role.
- A record that documents who was trained, on what, when and with what result.
It is the document you show when someone asks.
The four parts of the record
- Syllabus by role
- What is taught to each profile and in how much depth.
- Attendance register
- Who attended, which session and on what date.
- Assessment results
- What was measured in each role and with what result.
- Version control
- Which version of each piece of content was delivered, and when.
It is handed over at the end of the programme and updated with each edition.
What has already happened, with sources
An analysis of the four rulings (2024-2026), the common pattern that explains them and a ten-question self-check list for your firm. Every statement, with its source cited.
Read the report (in Spanish) → · Download the PDF (in Spanish)
Law firms, tax and employment advisory firms and gestorías
I work with law firms, tax and employment advisory firms and gestorías (administrative-services firms) of between 15 and 150 professionals in Spain and Latin America.
- AI is already used every day on work with legal effects, whether it has been authorised or not.
- The risk stopped being theoretical sooner than in any other sector: the sanctions exist, are published and are increasing.
- Someone has to answer for it (the managing partner or the management) and today they have nothing to answer with.
If your organisation fits that profile, the diagnostic is the starting point. Who is behind it: here (in Spanish).
Three steps, with no prior commitment
-
Questionnaire
A questionnaire on the current use of AI in your organisation. It takes about 20 minutes to answer. Free.
-
Conversation
A 30-minute conversation to go over the results and the preliminary map of use and risk by role.
-
Proposal
If it makes sense to continue, a fixed proposal of learning paths, schedule and price. You keep the map, whether you hire me or not.
A full team programme falls within €4,000-8,000, depending on the number of role learning paths, the size of the workforce and the depth of the assessment. I publish the range because asking for a quote shouldn't be an act of faith: what moves the price.
Ongoing training
Keeping the record alive
€190/monthno lock-inTraining for new joiners, regulatory updates and a quarterly remote session. Your record of the measures taken stays up to date instead of being frozen at the date of the programme.
What we are asked before starting
Does this really apply to us?
Yes, as an obligation of means. Article 4 of Regulation (EU) 2024/1689 has applied since 2 February 2025 and requires every organisation that uses AI systems (with no minimum size threshold) to take measures to support the AI literacy of its staff, taking into account their technical knowledge, experience, education and training and the context of use. It is an obligation of means: the article itself states that it does not require any specific level of AI literacy to be achieved by any particular person. Regulation (EU) 2026/1744 of 8 July 2026 (the Digital Omnibus) rewrote the article in that sense and did not postpone the obligation.
Source: Regulation (EU) 2024/1689, Article 4, consolidated version of 27 July 2026 on EUR-Lex
Is it enough that my team has taken an online course?
It can be one of the measures, but the article asks for each person's knowledge, experience and training, and the context in which they use AI, to be taken into account. An identical course for the whole workforce captures none of that; role-based learning paths do, and the record documents the measures taken (syllabus, attendance and assessment). The rule sets neither a format nor a minimum level.
Source: Regulation (EU) 2024/1689, Article 4, consolidated version of 27 July 2026 on EUR-Lex
Is there an official accreditation?
No, and it is worth saying clearly: the completion records issued by this training are our own. There is currently no official Article 4 accreditation and no public body that issues one, so no training can offer it. The rule does not ask for a seal or for a level per person: it asks for proportionate measures to be taken, and it is advisable to be able to document them; that record is the file we hand over.
How much does it cost?
A full team programme falls within €4,000-8,000, depending on the number of role learning paths, the size of the workforce and the depth of the assessment. What moves the price is published on the method page; I prefer the range to be known before the first conversation.
How much time will it take my staff?
It depends on the role, and it is sized in the diagnostic. As a usual reference: the general literacy foundation is around 20 hours per person; a role learning path, between 30 and 40 hours; the governance programme on Regulation (EU) 2024/1689 (the AI Act) for management and compliance, about 20 hours.
What if we already use AI without any control?
That is the normal case, not the exception: in the four court rulings analysed in the report (in Spanish), the problem was precisely the unsupervised use of general-purpose tools. That is why the first step is the free AI use questionnaire: knowing who uses what, for which task and with what risk, before training anyone.