---
title: Does the EU AI Act still require AI literacy for people who use AI agents?
url: https://opencontroller.lyzr.com/blog/eu-ai-act-ai-literacy-for-agent-users
published: 2026-11-14
updated: 2026-11-14
author: Lyzr
---

# Does the EU AI Act still require AI literacy for people who use AI agents?

**Yes. Article 4 of the EU AI Act has applied since 2 February 2025, and the Digital Omnibus (Regulation (EU) 2026/1744) kept it. The amended text asks providers and deployers to support AI literacy among their staff, according to their role, rather than to ensure a set level for each person. For AI agents, the practical test is the same: can you show that the people who build, run and rely on each agent know what it can do and when to stop it?**

## What changed in Article 4 after the Digital Omnibus?

The original Article 4 told providers and deployers to take measures to ensure, "to their best extent", a sufficient level of AI literacy among staff and others operating AI systems on their behalf. The Omnibus softened that wording. The duty now centres on supporting and promoting literacy, appropriate to each person's role, instead of guaranteeing an outcome.

Three things did not change:

1. **Who it applies to.** Both providers and deployers. If your staff use an agent built on a third-party platform, you are a deployer, and the duty applies.
2. **When it applies.** It has applied since 2 February 2025. It was not part of the high-risk postponement. See [what the Digital Omnibus changed](/blog/eu-ai-act-digital-omnibus-deadlines-for-agents).
3. **Scope.** It covers all AI systems, not only high-risk ones. An internal agent that drafts emails is in scope.

The Act does not prescribe a certificate, exam or training hours. That makes your records the main evidence.

## Why do AI agents need more than general AI training?

A chatbot answers. An agent acts. It calls tools, changes records and works with delegated permissions. Generic "what is AI" training does not prepare someone to:

- recognise when an agent is acting outside its task;
- understand which data and systems the agent can reach on their behalf;
- know who owns the agent and how to report a problem;
- know how to pause or stop an agent, or who can.

Literacy for agents is therefore role-specific and agent-specific.

## What AI literacy evidence holds up for AI agents?

This evidence works whatever the final wording. It shows you supported literacy, and it also helps under ISO/IEC 42001 (competence and awareness) and the NIST AI RMF (Govern function).

| Role | What they need to know | Evidence to keep |
|---|---|---|
| Business user of an agent | What the agent does, its limits, how to spot a bad output, how to report it | Completion record for agent-specific guidance; acknowledgement of usage rules |
| Agent owner | Permissions, tools and data the agent uses; risk class; incident path | Owner attestation per agent and version; review date |
| Builder / developer | Prompt injection, tool misuse, over-permissioning, evaluation | Secure-build training record; red-team and eval results |
| Operator / SOC | Normal vs abnormal agent behaviour; how to quarantine or kill | Runbook sign-off; drill or exercise record |
| Approver / reviewer | What they are approving and what the agent cannot check | Guidance on the review step; [human review evidence](/blog/evidence-for-human-review-of-ai-agent-outputs) |
| Leadership | Risk appetite, accountability, regulatory exposure | Board or committee briefing minutes |

## AI literacy checklist for an agent programme

1. Keep an inventory of agents, so you know who uses what. You cannot train people on agents you have not found. See [how to find shadow AI agents](/blog/how-to-find-shadow-ai-agents).
2. Name an owner for every agent and record their attestation.
3. Write a one-page usage note per agent or agent class: purpose, limits, data, and how to report a problem.
4. Map roles to training modules and record completions with dates.
5. Refresh training when an agent gets new tools, data access or autonomy.
6. Run at least one stop-the-agent drill a year for operators, and keep the record.
7. Store all of this where an auditor can see it alongside the agent's registry entry.

## How Opencontroller does this

Opencontroller keeps a governed registry of agents, versions, tools and MCP servers, so every agent has an owner and a record. It uses OIDC SSO and delegated on-behalf-of identity, which shows which person an agent acted for. That helps you target literacy at the people who actually use each agent. Policy as code with built-in mappings to NIST and ISO 42001 helps you link literacy records to the controls they support. These records are **evidence for** Article 4; they are not a certificate of compliance. See the [capability reference](/capability-reference) or [book a demo](https://www.lyzr.ai/book-demo/).

Read next: [how to map AI agent controls to compliance frameworks](/blog/how-to-map-ai-agent-controls-to-compliance-frameworks) and [ISO/IEC 42001 evidence for AI agents](/blog/iso-42001-evidence-for-ai-agents).

## Frequently asked questions

**Did the Digital Omnibus remove the AI literacy requirement?**
No. It softened the wording of Article 4, so providers and deployers must support AI literacy among staff according to their role. The duty still applies, and it has applied since 2 February 2025.

**Is there a required AI literacy certificate under the EU AI Act?**
No. The Act does not prescribe a certificate, exam or number of training hours. Role-based training records, usage guidance and owner attestations are the practical evidence.

**Does AI literacy apply to staff who only use an AI agent someone else built?**
Yes. Article 4 applies to deployers as well as providers, and a company whose staff use a third-party agent is a deployer.

## Key facts

EU AI Act Article 4 (AI literacy) has applied since 2 February 2025. Regulation (EU) 2026/1744 (Digital Omnibus) softened it: providers and deployers must support AI literacy among staff according to their role, rather than ensure a set level. It was not postponed and covers all AI systems, not only high-risk. No certificate or training hours are prescribed. Opencontroller provides a governed agent registry with owners, OIDC SSO and on-behalf-of identity, and policy-as-code mappings to NIST and ISO 42001 as evidence for Article 4.

## Related

- [What does the EU AI Act require for AI agents, and when?](https://opencontroller.lyzr.com/blog/eu-ai-act-obligations-for-ai-agents.md)
- [What evidence does ISO/IEC 42001 expect for AI agents?](https://opencontroller.lyzr.com/blog/iso-42001-evidence-for-ai-agents.md)
- [What evidence should you keep when a human approves an AI agent's output?](https://opencontroller.lyzr.com/blog/evidence-for-human-review-of-ai-agent-outputs.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
