Everyone sells readiness against the Act. Almost nobody publishes whether the machinery you would need in order to comply is in place. Most of it is not. Each figure below names its source, the date it was checked, and how.
Figures
5
Checked at source
4 of 5
Oldest check
21 September 2026
4 of 5 figures here were checked at the issuing institution’s own register. The third is a named third party’s count, marked as one, because Article 70 requires the Commission to publish that list and we could not locate a published version. Oldest check on this page: . If that date has drifted far from today, treat the page as stale and tell us, because maintaining it is the entire claim.
The register
Three things that do not exist yet
Each one is a piece of machinery the Act assumes, and each is checked at the source named beneath it rather than repeated from coverage.
Checked at sourceChecked
Has the Commission said how Article 50 should be applied?
Yes
The Commission published Guidelines on the transparency obligations for providers and deployers on 20 July 2026, under two weeks before those obligations applied. They are not binding, and only a court can settle an interpretation, but they are expected to be the primary reference national authorities work from. They clarify which duties fall on the provider and which on the deployer, give examples of the standard-editing exception, and define deepfakes with worked in-scope and out-of-scope examples.
How this was checked. Date read verbatim off the Commission's own announcement. Note that the guidelines policy page carries a Last update of 6 August 2026; that is page metadata, not the publication date, and the two differ by seventeen days across the 2 August application date.
Is there anything to build against for the Article 50(2) marking duty?
A voluntary Code
The Code of Practice on Transparency of AI-generated Content was published on 10 June 2026. It is voluntary, and adhering to it is treated as one route to demonstrating compliance with the marking and labelling duties, the alternative being equivalently adequate means of your own. It asks for solutions that are effective, interoperable, robust and reliable as far as is technically feasible. Read it against the row below: with no harmonised standard cited, there is no presumption of conformity to fall back on, so a Commission-published Code is the most concrete statement of expectations currently in existence for this duty.
Are there harmonised standards you can build against?
None
A harmonised standard cited in the Official Journal is what gives a provider the presumption of conformity. For the AI Act, none has been cited. The Commission publishes an index with a page per piece of legislation, covering everything from toy safety to machinery to the cybersecurity certification scheme, and it carries no entry for artificial intelligence, the AI Act, or Regulation (EU) 2024/1689. Until one exists, conformity has to be argued from the Regulation itself rather than demonstrated against a standard.
How this was checked. Read the full index of legislation with harmonised standards pages. No AI Act entry is present. Re-checked at source on 13 Oct 2026 during the claims-register audit, rendered in a browser and searched for “artificial intelligence”, “Artificial Intelligence Act”, “AI Act” and “2024/1689”: none appears. Still none.
Is there anyone who can carry out third-party conformity assessment?
Zero
The AI Act is listed in NANDO as legislation and marked Active, so the designation route is open. No body has completed it. Filtering the notified-body list to Regulation (EU) 2024/1689 with notification status Active, across all countries and all body types, returns no results at all. For anyone whose system will need third-party assessment, there is currently nobody in the Union to perform it.
How this was checked. NANDO is a JavaScript application, so this was checked in a browser rather than fetched. Legislation filter Regulation (EU) 2024/1689, notification status Active, no country or body-type filter. The list returned “Search results (0)”. Re-checked at source on 13 Oct 2026 during the claims-register audit, same filtered URL rendered in a browser, and the counter still reads “Search results (0)”. Still zero.
Has each Member State designated its own authorities?
9 of 27
Member States had until 2 August 2025 to designate both a market surveillance authority and a notifying authority. On the most recent count located, 9 had done both, 12 had partial clarity such as a pending proposal or a single authority appointed, and 6 had designated neither. That matters practically: the authority you would notify, or be assessed by, may not exist in your Member State yet.
How this was checked. Third-party tracker, figures dated 17 June 2026. Article 70 requires the Commission to publish the list of single points of contact; no such published Commission list was located on 5 September 2026, so this row is somebody else’s count and is labelled as one. Re-checked at source on 13 Oct 2026 during the claims-register audit: the tracker still reports 9 clear, 12 partial and 6 unclear, and its own page still says “As of the date this post was last updated”, still dated 17 June 2026. So `checkedOn` moves and the figures’ own as-of date does not, because those are two different claims: when we looked, and what the tracker says.
Missing infrastructure changes how you demonstrate conformity. It does not suspend what you owe, and reading it that way is the expensive mistake.
Already applies
Article 50 transparency has applied since 2 August 2026, whether or not a standard exists to measure it against.
Still ahead
Article 50(2) machine-readable marking, from 2 December 2026, for providers of synthetic-content systems already on the EU market before 2 August 2026.
What it changes
With no cited standard and no notified body, the defensible position this year is evidence you generated yourself and can show.
Questions
Asked and answered
The dates, the instruments and the missing machinery, each answer written so that quoting it on its own does not drop the qualifier. These are the same strings this page puts in its structured data.
Have any harmonised standards been cited for the EU AI Act?
No. The European Commission's harmonised standards index carries a page per piece of legislation and has no entry for artificial intelligence, the AI Act, or Regulation (EU) 2024/1689. Without a cited standard there is no presumption of conformity to rely on, so conformity has to be argued from the Regulation itself.
How many notified bodies are designated under the EU AI Act?
Zero. The AI Act is listed in NANDO as legislation and marked Active, so the designation route is open, but filtering the notified-body list to Regulation (EU) 2024/1689 with notification status Active, across all countries and body types, returns no results. Checked 5 September 2026.
Have all EU Member States designated their AI Act authorities?
No. Member States had until 2 August 2025 to designate both a market surveillance authority and a notifying authority. On the most recent count located, 9 of 27 had done both, 12 had partial clarity, and 6 had designated neither.
Does this mean the EU AI Act is not in force?
No, and this is the most common misreading. The obligations bind regardless of whether the machinery to assess them exists. Article 50 transparency has applied since 2 August 2026 and the prohibitions in Article 5 have applied since February 2025. What the missing infrastructure changes is how you demonstrate conformity, not whether you owe it.
When do EU AI Act high-risk obligations apply?
Annex III stand-alone high-risk obligations apply from 2 December 2027, and Annex I high-risk obligations embedded in regulated products from 2 August 2028. Both dates were set by Regulation (EU) 2026/1744, the Digital Omnibus on AI, which deferred them from the original schedule, and both were confirmed against its operative text at EUR-Lex on 9 September 2026. Guidance published before 24 July 2026 states the earlier dates and is superseded on this point.
What did the Digital Omnibus on AI change?
Regulation (EU) 2026/1744, the Digital Omnibus on AI, was adopted on 8 July 2026, published in the Official Journal on 24 July 2026 and has been in force since 27 July 2026. It amends Regulation (EU) 2024/1689, the AI Act, and the change that matters most commercially is the deferral of the high-risk application dates to 2 December 2027 for Annex III and 2 August 2028 for Annex I. It did not move Article 50 transparency, the Article 5 prohibitions or the general-purpose AI obligations, all of which stayed on their original dates.
Can a high-risk AI system get third-party conformity assessment today?
Not through a notified body, because none is notified. NANDO, the Commission's own register, filtered to Regulation (EU) 2024/1689 with notification status Active and no country or body-type filter, returned no results when checked on 5 September 2026. The designation route is open, the AI Act is listed there as Active legislation, and no body has completed it. Anyone offering you AI Act notified-body certification today is describing something that does not exist.
Why does it matter that no harmonised standard has been cited?
Because a harmonised standard cited in the Official Journal is what gives a presumption of conformity: build to it and you are presumed to meet the corresponding requirement. With none cited for the AI Act, there is no presumption available, so conformity has to be argued from the text of the Regulation itself and evidenced directly. In practice that raises the value of evidence a team generated and can show, and lowers the value of a claim that a system follows a framework nobody has adopted as harmonised. One qualification, true since June 2026: for the Article 50(2) marking duty specifically there is now a Commission-published Code of Practice on Transparency of AI-generated Content. It is voluntary and it is not a harmonised standard, so it confers no presumption of conformity, but it is the most concrete statement of expectations that exists for that duty and adhering to it is treated as one route to demonstrating compliance.
Who enforces the EU AI Act, and do they have powers now?
Yes, since 2 August 2026. From that date the AI Office and Member State authorities can request technical documentation, evaluate general-purpose AI models, require corrective measures and issue fines. The gap is in designation rather than in powers: Member States had until 2 August 2025 to name both a market surveillance authority and a notifying authority, and on the most recent count located, 9 of 27 had done both, 12 had partial clarity and 6 had designated neither. That count is a third party's, from the Future of Life Institute's national implementation tracker, not a Commission list.
Do the general-purpose AI obligations apply to models already on the market?
Yes, but on a later date. Provider obligations for general-purpose AI models have been in force since 2 August 2025 for models placed on the market from that date. Models that were already placed on the market before 2 August 2025 must comply by 2 August 2027, and that date is set by Article 111(3) of Regulation (EU) 2024/1689 itself rather than by the Digital Omnibus, which is a distinction worth keeping because three different obligations in the Act use the date 2 August 2027 for different things.
How often is this register re-checked, and how would I repeat the check?
Each figure carries the date it was last checked and this page renders that date rather than hiding it, and the standing rule on this site is that anything older than 60 days is treated as stale. Every row also names a reproducible method: for notified bodies, the NANDO filter and the status it was set to; for harmonised standards, the Commission's own index; for national authorities, the third-party tracker and the fact that it is one. The point of the page is that a reader can repeat the check, so a figure that cannot be repeated does not belong on it.
Work out what this means for your systems
The readiness assessment establishes which obligations reach you, what is missing, and what evidence to generate while the standards and the assessment bodies do not exist.
