Skip to content
Amended. Regulation (EU) 2026/1744 entered into force 27 July 2026. See what moved →
EU AI Act ChecklistIndependent reference
Analysis · 3 August 2026

What actually changed on 2 August 2026

For two years this was the date the entire compliance industry pointed at. It arrived six days after the Digital Omnibus entered into force and turned out to be a real deadline for a different set of obligations than the one most organisations had been preparing for.

Art. 113 general application dateReg. (EU) 2026/1744

What took effect

  • Article 50 transparency. Disclosure that a person is interacting with an AI system; disclosure of emotion recognition and biometric categorisation to those exposed; marking of synthetic audio, image, video and text; deep fake disclosure. In detail →
  • Article 49 registration in the EU database. In detail →
  • National market surveillance authority powers, and with them the practical applicability of the Article 99 penalty regime to the obligations in force.
  • The general application date of the Regulation itself, under Article 113.

What did not

The Annex III high-risk obligations, risk management, data governance, technical documentation, logging, human oversight, accuracy, quality management, conformity assessment, post-market monitoring. Those moved to 2 December 2027 when Regulation (EU) 2026/1744 entered into force on 27 July 2026, six days before the old deadline.

The uncomfortable part

Expert analysis. A great deal of preparation was aimed at the wrong target. Organisations that spent 2025 and early 2026 building Annex IV documentation now have sixteen extra months for work they had nearly finished, and many of them arrived at 2 August with a chatbot that does not disclose it is a chatbot, which was the obligation that actually landed.

That is not a criticism of the preparation. It is an argument for tracking obligations by article rather than by headline date.

Where this usually goes next

Three situations account for most people reading this page. Each has a different answer.

A deal is blocked on an AI questionnaire

Legal will not sign until you can evidence how AI is governed. HumanAudit’s AI Trust Package is a fixed $3,500 over five business days: a public trust page, a pre-filled SIG Lite / CAIQ / SSPA Section K questionnaire bank, and your AI inventory and classification.

How this works for AI companies →

You need ISO/IEC 42001 documentation

23 clause-mapped AIMS documents with all 38 Annex A controls pre-populated, editable and yours to keep, from $199. Or score your gaps first: 18 questions, free, no signup to begin.

Free gap assessment →
See the three tiers →

You are not sure what reaches you

Twenty minutes with the founder. No prep, no deck, straight to the person accountable for the work. If none of this applies to you, you get told that on the call.

Book a free 20-minute call →

This reference is published by HumanAudit Inc. Not a law firm, not an accredited certification body, not a registered auditor. We build documentation, your counsel interprets it, and an accredited body of your choosing certifies you. How this is funded →

The next date is 2 December 2026, and it is not small

  1. Two new Article 5 prohibitions: non-consensual intimate imagery and CSAM generation. Unlike the rest of Article 5, these reach general-purpose generative products. Article 5 →
  2. Article 50(2) machine-readable marking for generative systems already on the market before 2 August 2026. Provenance marking is engineering work, not a policy update.

Four months. Both items have a design lead time longer than that if nothing has started.

A note on our own coverage

We retired our “15 weeks to August” countdown article rather than leave it standing, and it now returns a 410. The countdown it described has expired and the framing it used, that August was the high-risk deadline, stopped being true on 27 July 2026. Leaving expired urgency online to collect traffic is not something we are willing to do. Our editorial standards →

Next step

Check what is live for you, not what is in the news

The obligations that apply to you depend on what your systems do. Two organisations reading the same headline can have entirely different live obligation sets.

Not sure where you sit?

The classifier maps your system against Articles 5, 6, 50 and Annex III. Twelve questions, no email.

Run the classifier →

Questions

What EU AI Act obligations started on 2 August 2026?

Article 50 transparency obligations, Article 49 registration in the EU database, and the enforcement powers of national market surveillance authorities. 2 August 2026 was also the general application date of the Regulation. The Annex III high-risk obligations that were originally due on that date were deferred to 2 December 2027 by Regulation (EU) 2026/1744.

Is the EU AI Act being enforced now?

The legal basis for enforcement of the obligations that have applied is in place: national market surveillance authorities gained their powers on 2 August 2026 and the penalty regime applies to the obligations in force. Practical enforcement capacity varies by member state, and no public enforcement decisions under the AI Act had been reported at the time of writing. Absence of early decisions is not a safe basis for planning, since limitation periods and the Article 85 complaint right both operate regardless.