What "AI literacy" actually means.
A working definition for adults with jobs.
The short version
AI literacy is not a slogan. It is a working capacity. It means five things you can do: recognize when AI is in the room, use the common tools competently, judge their output, know what to keep out, and hold institutions accountable for how they use AI on you and on others. That is the definition. The rest of this memo shows what it looks like in different jobs, at different stages of life, and inside organizations of every size.
If you are a teacher, a doctor, a nurse, a public servant, a small business owner, a journalist, a lawyer, an accountant, an HR manager, a farmer, a parent, a board member, or anyone who has noticed that AI is now in your working life: this memo is for you. So is the foundation that wrote it.
This memo is a working definition, not a curriculum. Use it to evaluate any AI literacy claim, your own included.
Why this needs defining
"AI literacy" has become a phrase like "leadership" or "innovation". Everyone uses it. Almost nobody says what it means.
The EU AI Act made AI literacy a legal duty in February 2025 (see Memo 02). Companies are buying training. Schools are buying curricula. Ministries are writing strategies. Most of it is shaped around a definition that has never been written down in plain language.
A working definition matters because:
- A teacher needs to know whether a 90-minute webinar makes her staff "AI-literate".
- A clinic manager needs to know whether her nurses are ready before she lets an AI scribe into consultation rooms.
- A small business owner needs to know whether "I use ChatGPT every day" is enough.
- A board needs to know what the phrase means when it appears in their risk reports.
- A parent needs to know what to ask the school about how AI is being used on her child.
Without a definition, AI literacy becomes whatever a vendor sells. That is the problem this memo solves.
What AI literacy is not
Five misconceptions worth clearing first.
- It is not being a programmer. You do not need to write code. You do not need to understand the math. The bar is being literate, not being an engineer.
- It is not blind enthusiasm. Knowing how to praise AI is not literacy. Neither is knowing how to be afraid of it. Literacy is the capacity to make distinctions.
- It is not a certificate. A certificate proves you sat through training. Literacy proves you can do something afterwards. The two often have no relationship to each other.
- It is not the same as digital literacy. Digital literacy is using software well. AI literacy includes deciding when not to.
- It is not finished. The tools change. The capacity is a practice, not a state. People who are AI-literate today will need to keep being AI-literate next year.
The five capacities
An adult with working AI literacy can do five things. None of them require a degree. All of them require practice.
1. Recognition. Knowing when AI is in the room.
You can tell when a tool you use contains AI. The spell-checker that rephrases your sentences. The email assistant that summarizes threads. The HR system that ranks candidates. The imaging software that flags scans. The customer service chat that is not a human. The "smart" features inside Microsoft 365, Google Workspace, your CRM, your accounting software, your applicant tracking system.
Recognition is the first capacity because everything else depends on it. People cannot evaluate what they cannot see.
2. Use. Operating the common tools competently.
You can operate the everyday tools for your actual work. For most adults in 2026 that means ChatGPT, Microsoft Copilot, Google Gemini, and one or two domain-specific tools.
Competence here is concrete. You can write a useful prompt without thinking too hard. You can iterate. You can paste in a document and ask for a summary that reflects what the document actually says. You can ask follow-up questions in the same conversation. You can recognize a dead end and start over instead of arguing with the machine. You know that the way you ask shapes what you get back.
3. Evaluation. Judging when output is good, wrong, or dangerous.
You know that AI invents references. You know that AI confidently produces wrong dates, wrong names, wrong numbers. You know that AI defaults to plausible-sounding writing even when the underlying facts are missing.
You read AI output the way a good editor reads a draft. Useful, but not trusted. You verify the parts that matter. You can tell the difference between an AI summary that captured a document and one that smoothed it into something it never said.
4. Limits. Knowing what to keep out.
You know what not to put in. Patient records. Client data. Internal salaries. Draft contracts. Unpublished research. Personal data covered by GDPR. Anything subject to professional confidence.
You also know what not to delegate. Some decisions belong to humans: hiring and firing, diagnosis, sentencing, asylum, custody, child protection, life-and-death triage. You know the difference between AI as a draft assistant and AI as a decision-maker. You know where the line falls in your own work, and you do not let it drift.
5. Citizenship. Holding institutions accountable.
You can hold institutions accountable for how they use AI on you and on others. You can ask your bank why a loan was denied. You can ask your employer how the performance scoring works. You can read a privacy notice and find the questions that matter. You can be a parent, a citizen, a board member, a journalist, a voter who actually asks.
AI literacy is not only about your own use of these tools. It is about your standing in a society being remade by them.
Recognition, use, evaluation, limits, citizenship. Five capacities. That is the definition.
What it looks like in different jobs
The five capacities are constant. What changes is how they show up in the work.
A teacher who is AI-literate can spot AI-written student work without panicking. She designs assessments that AI cannot easily fake. She uses AI to draft lesson plans she then edits in her own voice. She can explain to a parent what AI did and did not contribute to a graded assignment. She does not pretend AI is a fad. She also does not let it write what she should still be teaching her students to write.
A doctor who is AI-literate uses an AI scribe for consultation notes while still checking the diagnosis line by hand. She knows which symptoms an AI triage tool tends to under-flag. She refuses to paste named patient histories into public chat tools. She can explain to a patient what role AI played in their care, in language the patient understands.
A nurse who is AI-literate can run an AI documentation tool quickly enough to save real time at the bedside. She knows when the medication-interaction AI is suggesting something dangerous. She can flag an automated risk score to the responsible doctor when the score does not match the patient in front of her.
A public servant who is AI-literate uses AI to draft routine correspondence in Danish and English. She knows which fields in a case file must never be uploaded to a public AI tool. She recognizes when an automated decision-support tool is biasing her colleagues. She writes case notes a human can defend in court if an AI recommendation is later challenged.
A small business owner who is AI-literate uses ChatGPT to write better marketing copy. She uses an AI bookkeeping tool with the same skepticism she gives her human bookkeeper. She has a one-page written policy for what her staff can and cannot put into AI tools. She knows whether the AI inside her hiring software is regulated under the EU AI Act.
A journalist who is AI-literate uses AI to read four-hundred-page reports quickly. She never quotes AI output without checking the original source. She can spot AI-generated images and audio in the wild. She writes about AI without falling for either the hype or the doom, because she actually understands what the tools can and cannot do.
A lawyer who is AI-literate uses AI to draft and to summarize. She never files an AI-generated brief without checking every citation, because she knows AI invents cases that do not exist. She advises clients on AI use without pretending to know more than she does.
An accountant who is AI-literate uses AI to categorize transactions and reconcile statements. She never lets AI sign off on a return. She knows which AI bookkeeping tools are GDPR-compliant for Danish clients. She can explain to a client what AI did and did not do in the preparation of their accounts.
An HR manager who is AI-literate knows when an AI hiring tool is a "high-risk system" under the EU AI Act. She never lets AI make the final shortlist. She can explain to a rejected candidate how AI was used in the process. She trains hiring managers to notice when AI scoring is drifting in ways nobody intended.
A farmer who is AI-literate uses AI imagery to spot crop disease early. She uses AI for paperwork and grant applications. She knows which of her data is being collected by the agricultural-tech vendor, where it is stored, and who else can see it. She does not trust an algorithm over what she can see in her own field.
Different jobs. Same five capacities. Applied differently.
What it looks like at different stages of life
AI literacy is not only a question of profession. It changes shape over a lifetime.
A student who is AI-literate uses AI to learn, not to substitute for learning. She can explain her own reasoning. She can do hard problems without the tool open. She knows what she has memorized and what she has outsourced. She knows the difference between using AI as a tutor and using it as a ghostwriter, and she chooses on purpose.
A young professional who is AI-literate uses AI to compress the boring parts of her job: meeting notes, status updates, first drafts. She knows which tasks are building her own skill and which can be safely delegated. She does not let AI write the things she should be learning to write herself.
A mid-career professional who is AI-literate has stopped pretending to dismiss AI. She has tried the tools. She has found two or three that genuinely help. She has also found a few that do not, and she has stopped using them. She knows when to push back on a manager who wants AI in places it does not belong.
A senior professional or leader who is AI-literate does not let consultants define AI for her organization. She has used the tools herself. She can read an AI vendor's pitch and ask the three or four questions that separate substance from theatre. She is honest with her staff about what AI will and will not do to their jobs.
A retiree who is AI-literate uses AI to read complex letters from the bank, the tax authority, or the hospital. She uses it for translation, for drafting replies, for understanding medication information. She knows AI sometimes invents details and reads carefully before acting. She does not give AI tools her bank details, her CPR number, or her medical records.
A parent who is AI-literate has a conversation with her children about AI before the school does. She knows which AI tools her teenager is actually using and what they are being used for. She knows the difference between an AI tutor and an AI companion app. She does not panic. She also does not pretend the tools are harmless.
What it looks like for organizations
An individual builds AI literacy through habit. An organization builds it through structure. Both matter. Neither is optional.
For a team
AI literacy for a team is a shared vocabulary. Everyone knows which tools are approved, what they are for, what they are not for, and where to ask questions. Decisions about AI are made in team meetings, not in private conversations with a vendor sales rep. New tools are not introduced through individual experiments that slowly become production. A literate team can say, out loud, what AI is doing in its work.
For a company or institution
AI literacy at this level is structural. There is a written policy that everyone has read. There is one person whose job description includes AI questions. New tools are tested in a small group before they are rolled out. AI use is on the agenda of the management meeting at least once a quarter. The board can answer the question: what is the worst that could realistically go wrong with our current AI use, and who is responsible if it happens.
For a leader
AI literacy is personal. You cannot delegate this. A director who has never used the tools herself cannot make good decisions about them. A municipal head who has never asked an AI to summarize a two-hundred-page report cannot judge what her staff are doing with it. A founder who only talks about AI in board meetings, but never opens ChatGPT, will buy the wrong tools and trust the wrong people. The leader's literacy sets the ceiling for the organization's.
For a board
AI literacy at board level means asking four questions every year:
- Which AI tools are in use across the organization, including the ones embedded in software we already pay for?
- What is the highest-risk use, and who is affected by it?
- Who is accountable, by name, if a decision made or supported by AI goes wrong?
- What are we doing to keep our staff literate as the tools change?
If a board cannot get clear answers to those four, the organization is not yet AI-literate. That is fine. Now you know what to fix.
How to know if you have it
Five honest tests, for individuals.
- Can you write a useful prompt without thinking about it?
- Can you tell when an AI tool is making something up?
- Can you name three things you will never put into a public AI tool, and why?
- Can you name one decision in your work that AI should not be making?
- Can you read a privacy notice and find the AI clauses?
Three tests, for teams.
- Does everyone know which tools are approved?
- Does everyone know who to ask if they are unsure?
- Has the team talked about AI together in the last thirty days?
Three tests, for organizations.
- Is there a one-page written policy that everyone has actually read?
- Has a senior person used the tools themselves in the last week?
- Can someone in the organization answer "what is our highest-risk AI use" in under five minutes?
If most answers are yes, you have working AI literacy. If most are no, this memo is your starting point. Either answer is fine. The honesty is what matters.
How to build it (without becoming a programmer)
Three short routes. None of them require code.
For an individual
- Use one general AI tool every day for two weeks. ChatGPT, Microsoft Copilot, or Google Gemini. Free versions are fine.
- Use it for real work, not toy questions.
- Keep a short list of when it helped and when it failed. One line each.
- After two weeks, read one short guide on writing prompts and one short guide on AI risks.
- After a month, write your own one-paragraph policy for yourself: what you will use AI for, what you will not, what you will never put in.
For a team
- One sixty-minute conversation about what tools the team is already using, in the open, without judgment.
- One short written policy. One page. Plain language.
- One person who answers questions when colleagues are unsure.
- One follow-up conversation in three months to update what was learned.
For an organization
- Inventory. List every AI tool in use, including the ones embedded in software you already pay for.
- Policy. One page. Written in plain language. Signed off by the senior leader.
- Brief. A thirty-minute staff session per team.
- Designate. Name one person as the contact for AI literacy questions.
- Review. Redo the inventory every six months, or sooner if a major new tool is introduced.
That is the practice. None of it requires being a programmer. All of it requires attention.
What the Foundation will do
The AI Literacy Foundation publishes plain-language guides, runs training cohorts for organizations that cannot afford a commercial provider, and reviews tools so that small Danish and European organizations can make informed choices before they deploy. Everything we produce is free at the point of access.
We work with teachers, doctors, nurses, public servants, small business owners, journalists, lawyers, accountants, HR managers, farmers, parents, board members, and the leaders of small and medium organizations across Denmark and the rest of Europe.
We do not sell AI tools. We do not advise on which vendor to buy. We define what literacy means and help people get there.
If you want to be notified when the next memo is published, or if you want training for your organization, write to us.
Closing
"AI literacy" is a working capacity, not a slogan. It is five things you can do: recognize when AI is in the room, use the common tools competently, evaluate what they produce, keep the wrong things out, and hold institutions accountable for how they use these systems on you and on others.
Anyone can build it. Most adults need a few months of honest practice. Most teams need one conversation and one page of written policy. Most organizations need a written inventory and a leader who has actually used the tools.
The phrase will keep being misused. Vendors will keep selling certificates. Policymakers will keep writing strategies. None of that builds literacy. Practice does. Honest conversation does. Plain language does.
That is what this foundation is for.
Ali Al Mokdad
Founder, AI Literacy Foundation
This memo is a working definition, not a curriculum. Use it to evaluate any AI literacy claim, your own included. It is not legal advice. For obligations under the EU AI Act, see Memo 02.