TwinLadderArticle 4 asks for measures
Since 27 July 2026, Article 4 of the EU AI Act is a best-efforts duty: providers and deployers of AI take measures that support AI literacy, and can show them.
From result to means
Article 4 has applied since 2 February 2025. As adopted in Regulation (EU) 2024/1689, it asked providers and deployers of AI systems to take measures to ensure, to their best extent, a sufficient level of AI literacy among their staff and others operating AI on their behalf.
That read as a duty of result: a level to be reached.
Regulation (EU) 2026/1744, the Digital Omnibus on AI, keeps the article and changes its force. In force since 27 July 2026, it asks for measures that support the development of AI literacy, and it says that no specific level has to be guaranteed for any individual.
The measures are still weighed against people’s technical knowledge, experience, education and the context of use. A duty of means is discharged by the measures taken, and by being able to show them.
Providers and deployers of AI systems shall take measuresto ensure, to their best extent, a sufficient level ofto support the development of AI literacy …
… does not require providers or deployers to guarantee any specific level of AI literacy of any individual.
The readiness check keeps each measure, with its date, in one place.See the readiness check
The dates
Move along the AI Act’s calendar and read what applies on each date. It opens where things stand today.
Now it can be enforced
National market-surveillance authorities hold enforcement powers from 2 August 2026. The duty they enforce is the amended one: measures, and the record of them.
- Aug 2024
- Feb 2025
- Jul 2026
- Aug 2026
- Dec 2027
- Aug 2028
What counts as evidence
A record of measures, kept in your own files: an inventory of the AI systems in use, a needs assessment by role, an AI usage policy, training records for each group, board minutes that name an owner, a register of AI decision boundaries and a dated review.
What the Commission has said
- Internal records of the measures are enough. External certification is not required.
- Pointing staff to user manuals is generally not considered sufficient.
- There is no size threshold: the article applies from the first AI system an organisation uses.
From the Commission’s AI literacy questions and answers of May 2025, written before the 2026 amendment. Measures are proportionate to the systems used and the people affected.
The TwinLadder readiness check reads where each of these stands on the seven pillars, and keeps the dated reading as part of the record.
Phrase by phrase
Each part of the article, with the pillar of the Twin Ladder Standard that reads it and the evidence that shows a measure in place. Switch one on to see it.
- It covers everyone who provides or deploys AI.
Technical infrastructure. A complete inventory of the AI systems in use, each with a named role. The readiness check reads whether you hold one.
- It asks for measures, with an owner.
Ethical & responsible use. A documented programme with a budget and a responsible owner, recorded in board minutes.
- It is judged by what you can show.
Evidence & documentation. A documented needs assessment, an evidence portfolio and a periodic review. The readiness check keeps the dated reading, with each later reading beside it.
- It reaches contractors as well as staff.
Training & development. Training extended to contractors and others acting on your behalf, with AI-literacy clauses in their contracts.
- It fits training to each person’s background.
Training & development. A baseline reading of each group, and training set by role and seniority.
- It weighs the context of use.
Deployment competence. AI systems classified by context and impact, and training that reflects the real use cases. The readiness check returns the gaps, ranked.
The mapping comes from the Twin Ladder Standard’s published Article 4 analysis.Read the standard
Two ways to show your measures
The readiness check reads the organisation on the seven pillars and places it against the compliance floor of 52: a dated, comparable reading of the measures already in place, and of what is missing.
An online workshop is a measure too: live, up to 12 people per group, from €2,900 plus VAT.
A best-efforts duty sets a modest bar. The reason to clear it well belongs to the board: where judgment lives in the organisation, and how it is formed now that AI sits in daily work.
The record
The app
First readingThe readiness checkFree once →After thatIncluded in every planFrom €89 a month →Background
AppThe judgment register, built from the bookSee it →ReadAI in HR: Article 4 for people teams8 min →ReadThe literacy paradox10 min →Start the readiness check See the workshops
Prices in euro, plus VAT. See the plans.
Questions
If your organisation provides or deploys an AI system, yes. There is no size threshold: the duty starts with the first AI system in use.
Since 27 July 2026 it is a best-efforts duty, as amended by Regulation (EU) 2026/1744.
No. The Commission has said that internal records of the measures are enough and that external certification is not required.
That guidance dates from May 2025, before the amendment. Read the consolidated text before relying on it.
Things you can show: an inventory of the AI systems in use, a needs assessment by role, a usage policy, training records, board minutes that name an owner and a dated review.
Pointing staff to user manuals is generally not considered enough.
National market-surveillance authorities, with enforcement powers from 2 August 2026. They enforce the amended duty: measures, and the record of them.
It is the floor. The question a board has to answer is larger: can your people still tell when the AI is wrong, and who will be able to in ten years?
That is judgment capital. Judgment Capital, the book, sets the method; the app keeps the register.
The readiness check is free once, and included in every plan after. Plans are per organisation, from €89 a month for up to 50 people. Online workshops cost €2,900 to €7,900 per group of up to 12. All plus VAT.
Put it on the record
The readiness check is a first reading of the seven pillars, dated and kept. It is free once, and included in every plan after.
Article 4 sets a modest bar. What a board builds above it, its judgment capital, is the subject of the book: Judgment Capital, the method TwinLadder works from.
Rather talk first? Write to us.