
Disclosure: NOXI publishes this page, and nothing on it is scored or ranked, NOXI included. The rule text, the dates and the slide order come from EUR-Lex, the European Commission's AI Act pages, the AI Office of Ireland, the UAE AI Office and SDAIA, each opened on 18 November 2026. NOXI appears only in a short build box, one router line and one FAQ answer.
AI literacy training under the EU AI Act: since the Digital Omnibus took effect on 27 July 2026, Article 4 asks providers and deployers to "take measures to support the development of AI literacy" of their staff, with no set level and no certificate. A 10-slide, 30-minute session and an internal record fit that. Checked 18 November 2026.
What this page is. A free, slide-by-slide staff deck for the EU AI Act's AI literacy rule, written for companies in the EU (Ireland, Germany, France, the Netherlands and the rest of the Union), for UK and US companies whose AI systems or AI output reach the EU, and with one short section on where UAE and Saudi guidance stands. It prints Article 4 as EUR-Lex serves it today next to the 2024 wording, then gives you the ten slides, a 30-minute plan, a record template and the dates to 2028.
How we checked. This is a documented review of official pages, not a new hands-on test. EUR-Lex, the Commission's AI Act pages, the AI Office of Ireland, the UAE AI Office and SDAIA were opened between 11:39 and 11:46 UTC on 18 November 2026; quotes are exact, the rest paraphrased. This page is not legal advice: the AI Act says what it says, your national authority enforces it, and your company's own AI policy decides how you apply it.
In short
- The rule nowArticle 4 of the AI Act, as replaced by Regulation (EU) 2026/1744 of 8 July 2026: providers and deployers "take measures to support" the AI literacy of staff and others who use AI on their behalf.
- What wentThe 2024 duty to "ensure, to their best extent, a sufficient level" of AI literacy. No specific level is required of any individual.
- Since whenArticle 4 has applied since 2 February 2025. National market surveillance authorities supervise and enforce it from 2 August 2026.
- ProofNo certificate; an internal record is enough, says the Commission.
- The deckTen slides in 30 minutes, from why we train to who to ask, with a case exercise and a quiz.
- CheckedOfficial pages only, signed out, 18 November 2026.
Key numbers
Facts for the EU, and for UK and US teams: the AI literacy rule as served on 18 November 2026
Eight facts carry the session, each from the official page that states it, with the time we opened it.
| Fact | What the official page says | Source | Read on |
|---|---|---|---|
| The rule today | Providers and deployers "shall take measures to support the development of AI literacy" of their staff | EUR-Lex, Regulation (EU) 2026/1744, point replacing Article 4 [1] | 18 Nov 2026, 11:40 UTC |
| No set level | The duty "does not require" a guarantee of "any specific level of AI literacy of any individual" | Same text; AI Act Service Desk, Article 4 [1] [4] | 18 Nov 2026, 11:39 UTC |
| The amending act | Regulation (EU) 2026/1744 of 8 July 2026, Official Journal L of 24 July 2026, in force on the third day after publication | EUR-Lex, OJ L 2026/1744 [1]; Commission AI Act page, updated 19 October 2026 [7] | 18 Nov 2026, 11:40 UTC |
| Start date | Chapters I and II, which include Article 4, apply from 2 February 2025 | EUR-Lex consolidated AI Act, Article 113(a) [2] | 18 Nov 2026, 11:40 UTC |
| Enforcement | National market surveillance authorities supervise and enforce Article 4 from 2 August 2026 | Commission, AI Literacy Questions and Answers, updated 20 October 2026 [6] | 18 Nov 2026, 11:41 UTC |
| Proof | "There is no need for a certificate"; an internal record of trainings is enough | Same Q&A [6] | 18 Nov 2026, 11:41 UTC |
| Outside the EU | The Act covers providers and deployers in a third country "where the output produced by the AI system is used in the Union" | EUR-Lex consolidated AI Act, Article 2(1)(c) [2] | 18 Nov 2026, 11:40 UTC |
| Ireland | A distributed model built on sectoral regulators, coordinated by the AI Office of Ireland, "established on 2 August 2026" under the Regulation of Artificial Intelligence Act 2026 | Department of Enterprise page [10]; AI Office of Ireland [11] | 18 Nov 2026, 11:42 to 11:44 UTC |
EU AI Act AI literacy requirements: what Article 4 says now, and what it used to say
The EU AI Act's AI literacy requirement changed in July 2026. Regulation (EU) 2026/1744, the Digital Omnibus on AI, says plainly that "Article 4 is replaced by the following", and the new text swaps a duty to reach a sufficient level for a duty to take supporting measures [1]. If your training slides or your policy still quote the 2024 wording, they are out of date.
Article 4 AI literacy: the 2024 wording and the July 2026 wording side by side
| Part | 2024 wording (Regulation (EU) 2024/1689 as first adopted) | Wording today (as replaced by Regulation (EU) 2026/1744) | What it means for your session |
|---|---|---|---|
| The duty | "take measures to ensure, to their best extent, a sufficient level of AI literacy" [3] | "take measures to support the development of AI literacy" [1] | Show the measures you took; nobody has to pass a test |
| Who | their staff and other persons dealing with the operation and use of AI systems on their behalf [3] | unchanged [1] | Contractors, service providers and sometimes clients count, per the Commission [6] |
| What to take into account | technical knowledge, experience, education and training, the context of use, the persons the systems are used on [3] | unchanged [1] | Different depth for different roles is allowed |
| Level | "sufficient level" [3] | no specific level of any individual has to be guaranteed [1] | No pass mark is required by law; your policy may still set one |
| Public support | not in Article 4 | Commission and Member States support providers and deployers, in particular SMEs, and the Commission publishes practical examples [1] | Watch the AI Act Service Desk for those examples |
| The Board | not in Article 4 | The AI Board adopts recommendations with common objectives, taking European competence frameworks into account [1] | A future update to your deck, once published |
The full first paragraph as EUR-Lex serves it today. "Providers and deployers of AI systems 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 and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used. This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual." [1] [2]
What "AI literacy" means. Article 3(56) defines it as the skills, knowledge and understanding that let providers, deployers and affected persons make an informed deployment of AI systems and gain awareness of the opportunities, the risks and the possible harm [2]. That definition was not changed, and it is the best one-line answer to "what are we training for".
Why the rule changed. The Omnibus's recital says a stringent duty to ensure a sufficient level did not suit all providers and deployers and burdened smaller enterprises in particular [1].
One date wrinkle. The Commission's AI Act page gives 27 July 2026 [7], which matches publication on 24 July plus three days [1]; its Q&A and repository page say "mid-July 2026" [6] [8]. Use 27 July 2026.
Training is still the normal route. The Commission's Q&A says that in many cases simply relying on a system's instructions for use, or asking staff to read them, "might be ineffective" [6]. For high-risk systems, Article 26(2) still requires oversight by people with "the necessary competence, training and authority" [2].
Who must train under the EU AI Act: providers, deployers and "other persons"
Nearly every company that lets staff use an AI tool for work is a deployer, so Article 4 reaches far beyond AI developers. The AI Act defines a deployer as a person or body "using an AI system under its authority", except in a personal non-professional activity [2]. Staff using a chatbot to draft advertising text are in scope: the Commission's Q&A says they "should be informed about the specific risks, for example hallucination" [6].
- Staff. Anyone in the organisation directly dealing with an AI system, from the analyst writing prompts to the manager signing off a report drafted with AI [6].
- Other persons. The Commission reads "persons dealing with the operation and use of AI systems on behalf of" a company as people under its organisational remit who are not employees, for example a contractor, a service provider or a client [6].
- People with AI degrees. Normally AI literate, the Commission says, though they still need to know your own systems and risks [6].
- No AI officer required. No specific governance structure is mandated for Article 4: no AI officer, no AI board [6].
UK and US companies whose AI output is used in the EU
The AI Act is EU law, but its scope is not limited to EU companies. Article 2(1)(a) covers providers that place AI systems on the EU market "irrespective of whether those providers are established or located within the Union or in a third country", Article 2(1)(b) covers deployers located in the Union, and Article 2(1)(c) covers providers and deployers in a third country where the AI system's output is used in the Union [2]. The Commission's Q&A says the same holds for Article 4 when the system is placed on the Union market, used in the Union or its use affects people in the EU [6].
In practice that means a London agency with a Dublin team, or a US software company selling an AI feature to customers in Germany, may have Article 4 duties for the staff and contractors involved. Which entity is in scope, and which authority is competent, are questions for your counsel and your company's policy.
AI literacy presentation: the 10-slide staff deck, slide by slide
An AI literacy presentation for staff works best as ten slides in the order a new user meets the questions: why, what the rule says, which tools, what never to paste, how to check the answer, when to label, how data law fits, a real case, the dates, and who to ask. The table maps each slide to the text it rests on.
| # | Slide | What goes on it | Source, read 18 November 2026 |
|---|---|---|---|
| 1 | Why we train | The definition of AI literacy in one sentence, where your staff already use AI, and what the session will leave them able to do. | AI Act Article 3(56) [2]; Commission Q&A [6] |
| 2 | What Article 4 says now | The current wording, the 2024 wording struck through, the 27 July 2026 date, and "no certificate". | Regulation (EU) 2026/1744 [1]; Q&A [6] |
| 3 | Tools we allow | Your approved list: tool, approved use, account type, who owns it. Anything not on it needs a request first. | Your company's AI policy |
| 4 | What not to upload | Personal data, client confidential material, unreleased figures, credentials. Use the approved tool or remove the details. | AI Act Article 2(7) and the GDPR [2]; your policy |
| 5 | Checking AI output | Hallucinated facts, invented sources, wrong numbers, bias. The person who sends it owns it. | Commission Q&A on hallucination [6] |
| 6 | Labelling deepfakes | Article 50(4): disclose deepfakes; AI text on matters of public interest needs a label unless a person reviewed it and holds editorial responsibility. | AI Act Article 50(4) [2]; Code of Practice [9] |
| 7 | Data protection link | The AI Act does not replace the GDPR; personal data in a prompt is still personal data. | AI Act Article 2(7) [2] |
| 8 | Case exercise | One realistic case from your own work, three questions, five minutes in pairs, then a short quiz. | Built from slides 3 to 7 |
| 9 | Dates that matter | 2 February 2025, 27 July 2026, 2 August 2026, 2 December 2026, 2 December 2027, 2 August 2028. | AI Act Service Desk timeline [5]; Article 113 [2] |
| 10 | Who to ask | Named owner for AI questions, the request route for a new tool, where the policy lives, the next refresher date. | Your company's AI policy; record template below |

What to say on each slide of the AI literacy deck
Each slide gets one finding as its title and two or three minutes of talk. The lines are written for a fictional company, Harbourline Analytics, in Dublin; swap in your company facts and keep the legal facts as printed.
Slide 1. Why we train
Title: "We all use AI at work, so we all need to know how it fails." Start with what your people already do: drafting emails, summarising reports, building slides. Then give the AI Act's definition in plain words: the skills to use AI in an informed way and to know its risks and possible harm [2]. Close with the promise: in 30 minutes everyone knows the approved tools, what never goes into a prompt, how to check an answer and who to ask.
Slide 2. What Article 4 says now
Title: "The law asks us to support your AI literacy, not to certify it." Put the current first sentence of Article 4 on the slide, with the old phrase "sufficient level" struck through and the date 27 July 2026. Say it has applied since 2 February 2025, is enforced from 2 August 2026, and needs no certificate [1] [6]. Do not add a fine figure: the AI Act leaves penalties to national rules, and the Commission's Q&A says any sanction must be proportionate to the individual case [2] [6].
Slide 3. Tools we allow
Title: "Use the tools on this list, on work accounts, for the jobs listed." A four-column table works: tool, approved use, account, owner, plus one line on tools not on the list. This slide is yours, not the law's: the AI Act prints no list of approved tools, so your AI policy is the source.
Slide 4. What not to upload
Title: "If you would not email it to a stranger, do not paste it into a public chatbot." List four kinds of material: personal data, client confidential documents, unreleased figures, passwords and keys. Then the alternative: the approved tool on a company account, or the text with names and numbers removed. Our guide to whether ChatGPT trains on your data walks through the settings one chatbot offers, and the confidential material guide shows how to strip a file before you upload it.
Slide 5. Checking AI output
Title: "AI drafts; you sign." Show one real hallucination: a figure that looks right and is not. The Commission names hallucination as the risk staff must be told about [6]. Then a three-step check: every number against its source, every quote against the original, every name and date once more. Our page on AI hallucinations has more cases you can borrow for this slide.
Slide 6. Labelling deepfakes and public-interest text
Title: "If it looks real and is not, say so." Article 50(4) has applied since 2 August 2026: deployers of AI that makes deepfake images, audio or video must disclose that the content was artificially generated or manipulated, and AI text published to inform the public on matters of public interest needs the same disclosure unless it went through human review or editorial control and a person holds editorial responsibility [2]. The Commission's voluntary Code of Practice on marking and labelling, published 10 June 2026, comes with EU icons for labels [9].
Slide 7. Data protection link
Title: "The AI Act sits on top of the GDPR, not instead of it." Article 2(7) says Union data protection law applies to personal data processed in connection with the AI Act and that, apart from two narrow provisions (Articles 4a and 59), the Act does not affect the GDPR [2]. So data protection training still applies to every prompt. Our sister page, the GDPR training presentation, gives the 12-slide refresher to pair with this one.
Slide 8. Case exercise
Title: "Spot three problems in five minutes." The Harbourline case: an account manager pastes forty client complaint emails, with names and phone numbers, into a personal chatbot account and asks for a summary slide, which states that "68% of complaints concern delivery delays", a figure the emails do not support. Ask pairs what should not have gone into the prompt, what is unchecked, and what the policy says to do instead. Then the quiz below.
Slide 9. Dates that matter
Title: "The rule already applies; the high-risk dates are still ahead." A timeline from 2 February 2025 to 2 August 2028 from the AI Act Service Desk [5], with today marked, so staff see the rule is already in force.
Slide 10. Who to ask
Title: "One owner, one route, one date." The named person for AI questions, the route for asking to use a new tool, where the policy lives, the next refresher date. Finish on the record: attendance is logged, the evidence the Commission describes [6].
AI literacy training for employees: a 30-minute session plan
Thirty minutes is enough for a first AI literacy session for employees who use general tools. The Commission says no training format is mandatory [6], so these timings are a starting point.
| Minutes | Slides | What happens |
|---|---|---|
| 0 to 3 | 1 | Why we train; a show of hands on who used AI this week |
| 3 to 7 | 2 | What Article 4 says now, old wording against new |
| 7 to 11 | 3 | The approved tools list, with one question from the room |
| 11 to 15 | 4 | What never goes into a prompt, and what to do instead |
| 15 to 19 | 5 | One live hallucination and the three-step check |
| 19 to 21 | 6 | Deepfake and public-interest text labels |
| 21 to 23 | 7 | The GDPR still applies |
| 23 to 28 | 8 | Case exercise in pairs, then the five-question quiz |
| 28 to 30 | 9 and 10 | Dates, who to ask, sign the attendance record |
The five quiz questions. 1) Since 27 July 2026, does Article 4 require a certificate? (No.) 2) Name two kinds of material that never go into a public chatbot. 3) What are the three checks before you send AI-drafted text? 4) When must a deepfake be disclosed? (When a deployer uses AI to generate or manipulate it, under Article 50(4); evidently artistic, creative or satirical work only needs a disclosure that does not spoil the work.) 5) Who do you ask before using a new AI tool?

Different depth for different roles
Article 4 tells you to take technical knowledge, experience, education and the context of use into account, and the Commission confirms that training at different levels of detail can be appropriate [1] [6]. One session for all, then modules by role, is a common shape.
| Group | What they need beyond the 30 minutes | Why, in the official text |
|---|---|---|
| Everyone using general AI tools | The 10-slide session and a yearly refresher | Article 4(1) and the Q&A's chatbot example [1] [6] |
| Managers who sign off AI-drafted work | How to review output, when to ask for sources, when labelling applies | Article 50(4) on human review and editorial responsibility [2] |
| People who buy or build AI systems | Provider and deployer roles, risk level, the AI Act in outline | The Q&A's minimum steps: role, risk, then actions [6] |
| Staff overseeing high-risk systems | System-specific training with competence and authority to intervene | Article 26(2), unchanged by the Omnibus [2] [6] |
How to record AI literacy training: no certificate, one internal log
The Commission's answer to "do we need certificates" is short: "There is no need for a certificate", and organisations can keep an internal record of trainings and other guiding initiatives [6]. A log with seven fields is enough to show what you did and when you will do it again.
| Field | Example entry (fictional) |
|---|---|
| Date and format | 3 December 2026, 30 minutes, in person, Dublin office, recorded for remote staff |
| Deck version | AI literacy deck v1.2, Article 4 as replaced by Regulation (EU) 2026/1744 |
| Attendees and roles | 42 staff and 3 contractors, by team |
| Tools covered | The approved list on slide 3 |
| Quiz result | Average 4.3 of 5; nobody is required to pass, so the score is feedback, not a gate |
| Owner | Head of compliance |
| Next refresher | June 2027, or sooner if the AI Board's recommendations are published |
Keep the deck file with the log; together they answer an authority or a client who asks. The Commission's Q&A adds that a sanction may be more likely where there is proof of an incident caused by a lack of appropriate training and guidance [6].
EU AI Act training dates that matter, 2025 to 2028
Eight dates frame slide 9, from the AI Act Service Desk's timeline, which reflects the Omnibus [5], and Article 113 [2].
| Date | What applies | Source, read 18 November 2026 |
|---|---|---|
| 1 August 2024 | The AI Act enters into force | AI Act Service Desk timeline [5] |
| 2 February 2025 | General provisions, including definitions and AI literacy, and the first prohibitions apply | Timeline [5]; Article 113(a) [2] |
| 2 August 2025 | General-purpose AI rules apply; Member States designate authorities and adopt penalty laws | Timeline [5] |
| 27 July 2026 | The Digital Omnibus on AI enters into force; Article 4 is replaced | Commission AI Act page [7]; OJ L 2026/1744 [1] |
| 2 August 2026 | Most rules apply, Article 50 transparency starts, enforcement starts for AI literacy | Timeline [5] |
| 2 December 2026 | New prohibitions on non-consensual sexual deepfakes and child sexual abuse material; transition ends for Article 50(2) marking by providers already on the market | Timeline [5]; Article 113(a) [2] |
| 2 December 2027 | Rules for high-risk AI systems in Annex III apply | Timeline [5]; Article 113(c) [2] |
| 2 August 2028 | Rules for high-risk AI in products covered by Annex I apply | Timeline [5]; Article 113(c) [2] |

AI literacy training in Ireland: who enforces it, and one page to read with care
Ireland enforces the AI Act through a distributed model built on its existing sectoral regulators, coordinated by the AI Office of Ireland [10] [11]. The Department of Enterprise's page says the Office "was established on 2 August 2026" as an independent statutory body under the Regulation of Artificial Intelligence Act 2026 [10]; the Office's own news page announced its first chief executive in an item dated 30 July 2026 [11]. Its site has sections for business ("Understanding the EU AI Act", obligations, a regulatory sandbox) and for the public, including "Education and literacy" resources [11].
A page to read with care. The Department of Enterprise's EU AI Act page, opened at 11:42 UTC on 18 November 2026, still lists the 2 February 2025 milestone as providers and deployers ensuring "a sufficient level of AI literacy", and still dates the Annex III high-risk rules to August 2026 and the Annex I rules to August 2027 [10]. These lines predate the Omnibus. For an Irish session, quote Article 4 from EUR-Lex or the AI Act Service Desk, and use the Department's page for what it is current on: the distributed model, the authorities and the AI Office's start date.
Irish staff of a UK parent may be covered by the same session, since Article 2 looks at where the AI is deployed and where its output is used.
Where UAE and Saudi guidance stands on AI literacy
Neither the UAE nor Saudi Arabia has an Article 4 style staff training duty on the official pages we read; both have published AI principles and guides that a training deck can draw on. For teams with offices in Dubai, Abu Dhabi or Riyadh, the EU slide on Article 4 is replaced by one slide on these documents, and your entity's own policy decides how they apply.
| Country | Document | What it says that fits a training deck | Source, read 18 November 2026 |
|---|---|---|---|
| UAE | The UAE Charter for the Development and Use of Artificial Intelligence (PDF on ai.gov.ae, uploaded July 2024) | Twelve principles, among them human oversight (6), transparency (5), data privacy (4) and number 11, "Promoting AI Awareness for an Inclusive Future" | UAE AI Office publications page and Charter PDF [12] |
| Saudi Arabia | SDAIA, Generative Artificial Intelligence Guidelines for Government (cover dated 2025) | Government employees must not enter data classified restricted or higher; all results should be human-reviewed; its checklist lists continuous training on GenAI benefits and risks | SDAIA Laws and Regulations page and PDF [13] |
| Saudi Arabia | SDAIA, Generative Artificial Intelligence Guidelines for Public (cover dated 2025) | Users should check generated content and follow laws such as the Personal Data Protection Law | Same page and PDF [13] |
For a Gulf slide. Title it "In the UAE and Saudi Arabia, we follow published AI guidance and our own policy." Put the UAE Charter's principle 11 and the SDAIA rule on classified data side by side, then your company's rule for personal data under the UAE's Federal Decree-Law No. 45 of 2021 or Saudi Arabia's Personal Data Protection Law (Royal Decree M/19), as your legal team words it. The SDAIA government guideline is written for government entities; private companies may use it as a reference.
AI literacy PPT: design rules that keep the deck short and checkable
An AI literacy PPT for staff should look like a briefing: one finding per slide title, one table or picture per slide, and every legal line sourced in the footer.
- Quote the law once, exactly. Article 4's first sentence on slide 2, with "Regulation (EU) 2026/1744" in the footer. Everything else in plain words.
- Strike through the old wording. Showing "sufficient level" crossed out is faster than explaining the amendment.
- Put a date on every slide. "Read 18 November 2026" in the footer tells the next presenter what to check again.
- Keep company facts and legal facts apart. Your tools and owners change monthly; the Article 4 text does not.
For general structure, our guide to building a training presentation covers openings, pacing and check-in questions that work for any internal course.
Which route fits your AI literacy session
Six routes, five of them other than NOXI, each with the official fact it rests on.
- If you want a ready self-paced course for everyone then Elements of AI, because its site says the free courses are made by MinnaLearn and the University of Helsinki, and it offers a 60-minute staff course it describes as mapped to the EU AI Act (read 18 November 2026, 11:44 UTC) [14].
- If you want a live quiz inside the session then Mentimeter, because its quiz page lists multiple-choice and true or false questions with a live leaderboard, and its plans page shows a free plan for 50 participants a month (served in euros, 11:45 UTC) [15].
- If you want the quiz to double as your training record then Google Forms, because its help page says you turn on "Make this a quiz", can collect email addresses and get summaries of every response (read 11:45 UTC) [16].
- If you want to see what other organisations do then the Commission's living repository of AI literacy practices, because the Commission says it holds more than 40 initiatives, from e-learning to bootcamps [6] [8].
- If you are a small company and want hands-on help then a European Digital Innovation Hub, because the Commission says more than 200 of them support SMEs and 9 in 10 already offer AI services, training included [6].
- If you want a designed 10-slide deck built from your own AI policy then NOXI's Presentation format, because you attach the policy as a PDF with one request, sources stay on the page, and it exports to PowerPoint [17].
Make it in NOXI: the 10-slide AI literacy deck from your AI policy
NOXI (noxi.ai) is an AI maker for presentations and documents: you describe what you need in one request, and it builds a designed deck or document in one of 14 formats that you can edit by chat and export to PowerPoint, Word or PDF; your first document is free, and after that a document costs $5, or Pro is $19 a month with $200 to spend.
This flow uses the Presentation format of NOXI 2.0: one request, your company's AI policy attached as a PDF, the ten slides above, then export to PowerPoint.
Make a 10-slide AI literacy training presentation for our Dublin office under Article 4 of the EU AI Act as replaced by Regulation (EU) 2026/1744. Take the approved tools, upload rules and owner from the attached AI policy. Slides: why we train, what Article 4 says now (current first sentence, old "sufficient level" struck through), tools we allow as a table, what not to upload, checking AI output, labelling deepfakes, the data protection link, a case exercise with a quiz, dates from 2 February 2025 to 2 August 2028, and who to ask. Cite EUR-Lex and the AI Act Service Desk, and put the date read in each footer.
On the tools slide, add a fifth column "account type" with "company licence" for every row, and change nothing else.
Step 1. Pick Presentation on Home
Open noxi.ai and create a free account; no card is needed for the first document. Pick the Presentation format, which starts as a 16:9 deck.
Step 2. Attach your AI policy as a PDF and send the request
Attach the policy as a PDF and paste Prompt 1. NOXI builds from your file and sketches every slide first (cover, key number, table, timeline, quote), then writes several slides at once.
Step 3. Keep the legal lines sourced
When a slide needs current facts, NOXI searches the live web first and every fact keeps its source on the page (inline link, footnote or sources list). Compare the Article 4 quote with the EUR-Lex text printed above before anyone presents it.
Step 4. Fix one slide by chat
Send Prompt 2, or select a slide and say what should change; NOXI changes only what you asked. NOXI also reads the deck through and fixes what does not agree, such as a date written two ways.
Step 5. Export to PowerPoint and present
Export to PowerPoint with editable text boxes and embedded fonts, or to PDF with clickable source links, and keep the file with your training log. You can also present full screen from NOXI or share a view link with remote staff.
What it costs in NOXI. The first document is free, once per account (up to 15 slides and 6 AI pictures, no card needed), so a 10-slide deck fits in it; its downloads carry a small Made with NOXI mark in the corner. After that every document costs $5 from a US-dollar balance, and editing in chat is free. Pro is $19 a month with $200 to spend every month. Any purchase, a top-up or Pro, removes the mark. Prices are in US dollars, tax not included. Your presentations, prompts and files are not used to train AI models.
Common mistakes in AI literacy training decks
Many AI literacy pages in search results on 18 November 2026 still quoted the 2024 rule, so check currency first. These six mistakes are the ones a compliance reviewer will spot first.
- Quoting "sufficient level" as the current law. That wording was replaced on 27 July 2026 [1].
- Promising a certificate. The Commission says none is needed [6]; a certificate from a course provider is a choice, not a legal requirement.
- Printing a fine for Article 4. Penalties come from national rules and must be proportionate [2] [6]. Do not borrow the fine levels for prohibited practices.
- Training only the AI team. Everyone who uses AI at work, and contractors acting for you, are in scope [6].
- Forgetting the GDPR. Article 2(7) keeps data protection law in force for every prompt [2].
- No record. Without the attendance log there is nothing to show [6].
What changed recently on these official pages
Five official pages behind this guide changed between June and October 2026; each is reflected above.
| Source | What moved | Date on the page | When we read it |
|---|---|---|---|
| EUR-Lex | Regulation (EU) 2026/1744 (Digital Omnibus on AI) replaced Article 4; the consolidated AI Act was updated | Regulation of 8 July 2026; OJ L 24 July 2026; consolidated text 27 July 2026 | 18 November 2026, 11:40 UTC |
| European Commission | AI Literacy Questions and Answers rewritten for the amended Article 4 | "Last update 20 October 2026" | 18 November 2026, 11:41 UTC |
| AI Office of Ireland | First chief executive appointed; Office established 2 August 2026, per the Department of Enterprise | news item dated 30 July 2026 | 18 November 2026, 11:44 UTC |
| European Commission | Living repository page updated to say Article 4 was amended | "Last update 10 August 2026" | 18 November 2026, 11:41 UTC |
| European Commission | Final Code of Practice on marking and labelling AI-generated content published | 10 June 2026, updated 12 June 2026 | 18 November 2026, 11:41 UTC |
What these sources don't measure
This is a structure guide built on the AI Act's text and the Commission's guidance. It leaves out several things that decide whether your own programme is right.
- Your national rules. Each Member State sets penalties and names its authorities; this page reads Ireland's pages only.
- US federal and state law. Only the EU AI Act's reach to US companies is covered; no US rule was reviewed.
- Whether a course is good. Elements of AI, Mentimeter and Google Forms appear for one documented fact each; none is rated, and replicating a repository practice does not by itself grant a presumption of compliance, the Commission says [6].
- High-risk systems. Article 26 training for high-risk deployers needs system-specific material this deck does not carry.
What would change this guide
Three official page changes would rewrite parts of this page, each tied to the page that would have to change and a date we would check again.
- 1. The AI Board adopts its Article 4(3) recommendations with common objectives, published on the Commission's AI literacy pages. Then slide 2, the role table and the session plan add those objectives. Re-check 18 December 2026.
- 2. The Commission publishes the Article 4(2) practical examples of compliance on the AI Act Service Desk. Then the record template and the router line on the living repository point to those examples instead. Re-check 18 December 2026.
- 3. Ireland's Department of Enterprise updates its EU AI Act page to the amended Article 4 wording and the 2 December 2027 date. Then the Ireland section drops its "read with care" note. Re-check 2 December 2026.
Frequently asked questions
- What is AI literacy training under the EU AI Act?
- It is the measures a provider or deployer takes so that staff and others using AI on its behalf understand the tools, the risks and the rules. Article 3(56) defines AI literacy as the skills, knowledge and understanding for informed use. A short session, role modules and a record of who attended are the usual shape.
- Is AI literacy training mandatory under Article 4?
- Taking measures is mandatory; a specific course is not. Since 27 July 2026, Article 4 requires providers and deployers to take measures to support staff AI literacy, with no set level. The Commission's Q&A says the format depends on context, though relying only on instructions for use may be ineffective. Checked 18 November 2026.
- What are the EU AI Act AI literacy requirements after July 2026?
- Article 4, as replaced by Regulation (EU) 2026/1744, asks providers and deployers to take measures supporting AI literacy of staff and others acting for them, considering their knowledge, experience and the context of use. No individual level must be guaranteed. Enforcement by national market surveillance authorities runs from 2 August 2026.
- Do we need a certificate to prove AI literacy training?
- No. The European Commission's AI literacy Q&A, updated 20 October 2026, says there is no need for a certificate and that organisations can keep an internal record of trainings and other guiding initiatives. Log the date, deck version, attendees, roles, tools covered and the next refresher date, and keep the deck with the log.
- How long should an AI literacy training session for employees be?
- About 30 minutes works for staff using general tools: ten slides covering the rule, approved tools, what not to upload, checking output, labels, data protection, a case exercise and who to ask. The Commission says no format is mandatory, so add role modules for managers, buyers of AI systems and high-risk overseers.
- Does Article 4 apply to UK and US companies?
- It can. Article 2 of the AI Act covers providers placing AI on the EU market wherever they are based, deployers located in the Union, and third-country providers and deployers whose AI output is used in the Union. The Commission says this holds for Article 4 too. Ask your counsel which entity is in scope.
- Who enforces AI literacy training in Ireland?
- Ireland uses a distributed model built on its sectoral regulators, coordinated by the AI Office of Ireland, which the Department of Enterprise says was established on 2 August 2026 under the Regulation of Artificial Intelligence Act 2026. That Department page still describes the 2024 Article 4 duty, so take the rule text from EUR-Lex.
- What should an AI literacy presentation cover?
- Ten slides: why we train, what Article 4 says now, tools we allow, what not to upload, checking AI output, labelling deepfakes under Article 50(4), the data protection link, a case exercise with a quiz, dates from 2 February 2025 to 2 August 2028, and who to ask. Source every legal line on the slide.
- Do UAE or Saudi rules require AI literacy training?
- The official pages read on 18 November 2026 show guidance rather than an Article 4 style duty. The UAE Charter for the Development and Use of AI lists promoting AI awareness as principle 11. SDAIA's 2025 generative AI guideline for government lists continuous staff training in its checklist. Your entity's own policy decides.
- Can NOXI make an AI literacy training presentation?
- Yes. In NOXI's Presentation format you attach your AI policy as a PDF and ask for the ten slides; facts keep their sources on the page, and you export to PowerPoint or PDF. The first document is free with a small Made with NOXI mark; after that a document costs $5.
Sources & methodology
Every page below was opened signed out on 18 November 2026; times are UTC. EUR-Lex and the Commission's pages were read in a real browser with a fresh profile. This is a documented review of official pages, not a new hands-on test, and not legal advice. Our sister guide to the EIC Accelerator pitch deck uses the same method for the EU's start-up funding rules.
- [1] EUR-Lex: Regulation (EU) 2026/1744 of 8 July 2026 (Digital Omnibus on AI), Official Journal L, 24 July 2026; recital on Article 4, point replacing Article 4, entry into force on the third day after publication; read 11:40 UTC, checked 18 November 2026.
- [2] EUR-Lex: Regulation (EU) 2024/1689, consolidated text of 27 July 2026; Articles 2(1), 2(7), 3(56), 26(2), 50, 99 and 113, and the definition of a deep fake; read 11:40 UTC, checked 18 November 2026.
- [3] EUR-Lex: Regulation (EU) 2024/1689 as published in 2024, Article 4 in its original wording; read 11:40 UTC, checked 18 November 2026.
- [4] AI Act Service Desk: Article 4: AI literacy, marked AMENDED, based on the consolidated version as at 27 July 2026; read 11:39 UTC, checked 18 November 2026.
- [5] AI Act Service Desk: Timeline for the Implementation of the EU AI Act, reflecting the Digital Omnibus; read 11:41 UTC, checked 18 November 2026.
- [6] European Commission: AI Literacy, Questions and Answers, "Last update 20 October 2026"; read 11:41 UTC, checked 18 November 2026.
- [7] European Commission: AI Act policy page, "Last update 19 October 2026", AI Omnibus in force on 27 July 2026; read 11:41 UTC, checked 18 November 2026.
- [8] European Commission: Living repository to foster learning and exchange on AI literacy, "Last update 10 August 2026"; read 11:41 UTC, checked 18 November 2026.
- [9] European Commission: Code of Practice on marking and labelling AI-generated content, published 10 June 2026, updated 12 June 2026; read 11:41 UTC, checked 18 November 2026.
- [10] Department of Enterprise, Tourism and Employment (Ireland): EU Artificial Intelligence (AI) Act, distributed model, milestone list and the AI Office's establishment date; read 11:42 UTC, checked 18 November 2026.
- [11] AI Office of Ireland: home page and news, item dated 30 July 2026 on its first chief executive; read 11:44 UTC, checked 18 November 2026.
- [12] UAE Artificial Intelligence Office: Publications and The UAE Charter for the Development and Use of Artificial Intelligence (PDF), twelve general principles; read 11:42 UTC, checked 18 November 2026.
- [13] Saudi Data and AI Authority (SDAIA): Laws and Regulations page, with the Generative Artificial Intelligence Guidelines for Government and for Public (both cover dated 2025); read 11:42 to 11:46 UTC, checked 18 November 2026.
- [14] Elements of AI: home page, MinnaLearn and the University of Helsinki, staff course offer; read 11:44 UTC, checked 18 November 2026.
- [15] Mentimeter: quiz feature page and plans, free plan for 50 participants a month, prices served in euros to a UAE address; read 11:45 UTC, checked 18 November 2026.
- [16] Google Docs Editors Help: Create and grade quizzes with Google Forms; read 11:45 UTC, checked 18 November 2026.
- [17] NOXI: NOXI 2.0 product record, rows A (formats), B (research with sources), C (plans before designing), F (self-checks), G (editing), I (export), J (present and share), N (price, from noxi.ai/pricing, read 18 November 2026), Q (Made with NOXI mark on the free document) and S (your own files as input), with the NOXI 2.0 launch article and the privacy policy; checked 18 November 2026.



