# AI literacy under Article 4 of the AI Act: what to do and what to record

URL: https://privacyboard.io/blog/ai-literacy-ai-act-article-4

Published 2026-11-18, updated 2026-11-18.

Article 4 of the EU AI Act has applied since 2 February 2025. Since the Digital Omnibus (Regulation (EU) 2026/1744), providers and deployers must take measures to support the AI literacy of their staff and others using AI on their behalf, with no specific level required. No certificate is needed, but you should keep a record of the measures.

Article 4 of the EU AI Act is the one duty that reaches almost every company using AI at work, including a team that only uses a chatbot to draft emails. Since the Digital Omnibus on AI, it asks providers and deployers to take measures to support the AI literacy of the people who use AI for them. It no longer asks you to ensure a "sufficient level", but it still asks you to act, and you need a way to show what you did.

This guide is for the person who answers for that: what the text says now, who it covers, who checks it, and what to keep on file. It reflects the texts and guidance as of 18 November 2026. It is general information, not legal advice.

## What does Article 4 say since the Digital Omnibus?

The original Article 4 of the AI Act (Regulation (EU) 2024/1689) asked providers and deployers to "take measures to ensure, to their best extent, a sufficient level of AI literacy" of their staff. It has applied since 2 February 2025.

The Digital Omnibus on AI, Regulation (EU) 2026/1744, was adopted on 8 July 2026, published in the Official Journal on 24 July 2026 and entered into force three days later. It replaced Article 4 with three paragraphs:

1. Providers and deployers "shall take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf", taking into account their technical knowledge, experience, education and training, the context of use, and the people the AI is used on. A new last sentence adds that the duty does not require them to secure any specific level of AI literacy for any individual.
2. The Commission and the Member States support providers and deployers, "in particular SMEs", and the Commission publishes practical examples on the AI Act's single information platform.
3. The AI Board adopts recommendations, drawing on European competence frameworks, with common objectives.

Recital 8 of the Omnibus explains why: stakeholders found the earlier wording too demanding for many organisations, and it added a burden for smaller companies.

| | Before the Omnibus | Since the Omnibus |
|---|---|---|
| Core duty | Take measures to ensure, to their best extent, a sufficient level of AI literacy | Take measures to support the development of AI literacy |
| Level required | A "sufficient level" | No specific level for any individual |
| Who is covered | Staff and other persons dealing with AI on their behalf | Same |
| Factors to consider | Knowledge, experience, education, training, context, people affected | Same |
| Public support | Not in the article | Commission and Member States support, in particular SMEs; Board recommendations |
| Applies since | 2 February 2025 | Same date; new wording in force since late July 2026 |

In short: the bar moved from a result to an effort. The effort is still mandatory.

## What counts as AI literacy?

Article 3(56) defines AI literacy as the "skills, knowledge and understanding" that let providers, deployers and affected persons make an informed use of AI systems and be aware of their opportunities, risks and possible harm.

The European Commission's AI literacy Q&A, last updated on 20 October 2026, suggests building measures around four points:

- **A general understanding of AI** in the organisation: what it is, how it works, which AI you use, and its opportunities and dangers.
- **Your role**: whether you provide AI systems or deploy systems made by others.
- **The risk** of the systems used: what staff need to know about it and how to reduce it.
- **The people and the context**: differences in technical knowledge, experience and training, the sector and the purpose of use.

The Q&A adds that legal and ethical aspects belong in the measures, and that simply relying on the instructions for use, or asking staff to read them, "might be ineffective".

## Does using ChatGPT at work bring you into scope?

Yes. The Commission's Q&A answers this directly: a company whose employees use a general chatbot to write advertising copy or translate texts is a deployer, and Article 4 applies. Staff should at least know the specific risks, such as hallucinations.

That makes Article 4 wider than most of the AI Act. A company with no AI in its product, no high-risk use and no customer-facing chatbot can still be covered because its sales or support team uses AI tools.

## Who must be covered, beyond employees?

The text says "staff and other persons dealing with the operation and use of AI systems on their behalf". According to the Commission's Q&A, these other persons are people "broadly under the organisational remit", such as a contractor, a service provider or a client.

In practice, think about:

- Freelancers and agencies who use AI on your account (content, translation, development).
- Outsourced support teams working in your helpdesk with its AI features switched on.
- Interns and temporary staff, who are easy to forget in a training plan.
- Clients, where the risk of the system you provide calls for guidance on its use.

You do not need to train a contractor's whole company. You need reasonable measures for the people who use AI on your behalf, such as a clause in the contract, your AI usage rules shared at onboarding, or access to your internal guide.

## Do you need training, a test or a certificate?

No certificate is needed. The Commission's Q&A says so in plain words, and adds that "no strict requirements or mandatory trainings are imposed". Germany's Bundesnetzagentur, which hosts the national AI service desk, says the same: organisations choose their own measures, internal or external, and formal training or certification is not required.

Measures that fit the text include:

| Measure | When it fits | Effort |
|---|---|---|
| One-page AI usage rules (which tools, what data never goes in, who to ask) | Every company using AI | A few hours to write, minutes to read |
| Short awareness session on how AI tools work and fail | Everyone using general AI tools | 30 to 60 minutes per group |
| Role-based training (prompting with care, checking outputs, personal data) | Teams using AI daily: support, sales, marketing, engineering | 1 to 3 hours |
| Product-specific guidance | Teams building AI features (providers) | Part of engineering onboarding |
| Human oversight training | Staff overseeing a high-risk system | Specific to the system |
| Clause or briefing for contractors | Agencies and freelancers using AI for you | One contract update |

The Commission keeps a "living repository" of more than 40 AI literacy practices from organisations of different sizes. Copying one of them does not by itself give a presumption of compliance with Article 4, according to the Q&A, but it is a useful starting point.

## How does this relate to high-risk systems?

For high-risk AI systems, the Omnibus did not lower the bar on people. Article 26(2) still requires deployers to "assign human oversight to natural persons who have the necessary competence, training and authority, as well as the necessary support". The Commission's Q&A confirms that this training duty remains. The high-risk rules for Annex III uses, such as recruitment or credit scoring, now apply from 2 December 2027.

So a company that will use AI to screen candidates has two layers: Article 4 for everyone using AI, and specific competence and training for the people overseeing that system.

## Who enforces Article 4, and what are the penalties?

National market surveillance authorities supervise and enforce the AI Act, not the European AI Office. According to the Commission's Q&A, they started supervising and enforcing the rules on 2 August 2026, while the duty itself has applied since 2 February 2025.

Article 4 is not in the lists of fixed maximum fines in Article 99(3) and (4) of the AI Act. Article 99(1) leaves it to each Member State to set the penalties and other enforcement measures, which may include fines, warnings and non-monetary measures. The amounts and the authority in charge therefore depend on the country; check the national law where you operate.

The Q&A also notes that the AI Act creates no right to compensation of its own, but a provider of an AI system can be sued under national law when harm occurs. The Bundesnetzagentur warns that a lack of AI literacy can be seen as a breach of a duty of care, especially if it leads to harm. A record of measures is your evidence either way.

## How does this fit with your GDPR work?

Most of the risk staff need to understand is about data: what they paste into an AI tool, where it goes, and whether the output about a person is correct. If you already run GDPR awareness training, add an AI part rather than building a separate track. Accountability under Article 5(2) and Article 24 of the GDPR already asks you to show the measures you take, and the same habit of recording dates, audiences and content serves both laws.

Your register of processing activities also tells you where to focus: the activities that use AI tools show which teams need role-based measures first.

## Checklist: an AI literacy programme you can show

- [ ] List the AI systems in use, by team, including general chatbots and AI features inside existing tools.
- [ ] For each system, note your role (provider or deployer) and the risk level.
- [ ] Identify the people covered: staff, contractors, agencies, temporary staff and, where relevant, clients.
- [ ] Write AI usage rules: approved tools, data that never goes in, checking outputs, who to ask.
- [ ] Run an awareness session for everyone who uses AI, with a short part on hallucinations and personal data.
- [ ] Add role-based measures for teams using AI daily or building AI features.
- [ ] Add AI rules to onboarding and to contracts with agencies and freelancers.
- [ ] Plan human oversight training for any future high-risk use.
- [ ] Record every measure with its date, audience and content.
- [ ] Review the programme at least once a year and when a new AI tool or feature arrives.

## What to document

- **AI inventory**: each AI system, the team using it, your role and its risk level.
- **Needs assessment**: for each group of people, what they need to know and why (knowledge, context, people affected).
- **Measures**: type (rules, session, training, guide, contract clause), content or materials, date, duration and trainer.
- **Participants**: who received each measure; attendance lists or acknowledgements for usage rules.
- **External people**: which contractors and agencies got the rules or a clause, and when.
- **Reviews**: the date of each yearly review, what changed and refresher sessions planned.
- **Guidance relied on**: the Commission Q&A version, national authority guidance, and any repository practice you adapted.

In Privacyboard, the AI Act tab under Risk checks classifies each activity that uses AI, and the company profile records each AI literacy measure with its kind, audience and date.

## FAQ

### Did the Digital Omnibus remove the AI literacy obligation?

No. It reworded it. Providers and deployers must still take measures to support the AI literacy of their staff and others using AI on their behalf. What changed is that they no longer have to ensure a "sufficient level", and the text now says no specific level is required for any individual.

### When did Article 4 start to apply?

On 2 February 2025, together with the prohibited practices. The new wording has applied since the Omnibus entered into force in late July 2026. National market surveillance authorities started supervising the rules on 2 August 2026, according to the Commission.

### Is there a fixed fine for not supporting AI literacy?

No fixed amount in the AI Act itself. Article 4 is not in the fine ranges of Article 99(3) and (4); each Member State sets the penalties and other enforcement measures. A lack of literacy can also weigh against you if an AI tool causes harm.

### Do we need to train our clients?

Not as a rule. The Commission's Q&A names clients among the persons who may be covered, depending on the risk of the system. For a low-risk feature, clear documentation is usually enough; for a riskier system, guidance on its use is worth adding.

### Is an e-learning certificate enough evidence?

It helps, but it is not required and it may not be enough on its own. What shows the effort is a record of measures matched to roles and risks: who got what, when, and why. A generic course with no link to the tools your people use is weaker evidence.

## Sources

- [AI Act, Regulation (EU) 2024/1689, EUR-Lex](https://eur-lex.europa.eu/eli/reg/2024/1689/oj)
- [Digital Omnibus on AI, Regulation (EU) 2026/1744, EUR-Lex](https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ%3AL_202601744)
- [European Commission: AI Literacy, Questions & Answers (updated 20 October 2026)](https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers)
- [European Commission AI Act Service Desk: Article 26, obligations of deployers of high-risk AI systems](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-26)
- [European Commission AI Act Service Desk: Article 99, penalties](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-99)
- [Bundesnetzagentur: AI literacy](https://bundesnetzagentur.de/EN/Areas/Digitalisation/AI/07_Literacy/start.html)
