08 / Trust & Governance
AI literacy
The working understanding people need to use AI well and safely: what it can do, where it fails, and when to check. Under the EU AI Act, organisations must make sure their staff have it.
How you’ll hear it
“What training do we actually need?”
“AI literacy first. People need to know when not to trust it.”
Avoid
Equating AI literacy with a list of prompt tips.
The longer version
AI literacy is the knowledge and judgement to use AI appropriately for your role: its strengths, its failure modes, the risks to data and to people, and the checks that apply. Article 4 of the EU AI Act makes it an obligation for organisations using AI. A shared vocabulary, like this one, is where it starts.
In depth
AI literacy is the working understanding people need to use AI appropriately for their role: what it is good at, where it fails, what data it should never see, and when its output must be checked. It is judgement, not technical skill.
Under Article 4 of the EU AI Act, organisations that use AI must take measures to ensure a sufficient level of AI literacy among the staff who work with it. That duty has applied since February 2025, and it applies whether or not a system counts as high risk.
Literacy differs by role. A customer service agent needs to recognise a confident wrong answer and know when to escalate. A procurement lead needs to ask the right questions about data and documentation. A board needs to understand enough to challenge a business case. One generic course rarely serves all three.
The cheapest place to start is shared vocabulary. When people mean the same thing by hallucination, agent or evals, the rest of the conversation gets easier.
What to do
- Decide what each role needs to know, rather than sending everyone on the same course.
- Teach failure modes first: hallucination, sycophancy and data leakage cause most real harm.
- Record who has been trained on what. You may be asked to show it.