16 more months for high-risk. Which duties still bind you now?

The AI Act amendment moved two dates and left the rest untouched. Prohibitions, GPAI duties and Article 50 transparency all apply today, delay or no delay.

Plan now EU

In short

  • The amendment approved on 16 June 2026 moved exactly two dates: Annex III high-risk obligations from 2 August 2026 to 2 December 2027, and Annex I from 2 August 2027 to 2 August 2028.
  • That is a deferral of 16 months for standalone high-risk systems and 12 months for AI embedded in regulated products. Nothing else in the timetable moved.
  • Article 50 transparency duties have applied since 2 August 2026. Generative systems already on the market got a short extension to 2 December 2026 for machine-readable marking, and only for that.
  • A new prohibition covering AI generated intimate imagery and child sexual abuse material carries its own date of 2 December 2026 and applies whether or not you run anything high risk.
  • Breaches of the transparency duties can reach EUR 15 million or 3% of worldwide annual turnover, so the duties that did not move are not the cheap ones.
Advertisement

What do AI assistants say about your organisation?

Model answers are becoming a channel your communications team does not control. EntityRise audits how assistants describe your company and whose name they give instead. From $19.

Audit your AI visibility at EntityRise.ai →

The deferral was reported as breathing room. It moved two dates and left everything else where it was.

If your programme paused on the strength of the headline, the duties that kept running are the ones with the nearer deadlines.

01What exactly did the amendment move?

Two dates, both for high-risk systems.

Obligations for standalone high-risk systems listed in Annex III moved from 2 August 2026 to 2 December 2027, a deferral of 16 months. Obligations for AI embedded in products already regulated under Annex I, such as medical devices, lifts and radio equipment, moved from 2 August 2027 to 2 August 2028, a deferral of 12 months.

The change was agreed provisionally on 7 May 2026 and approved by the European Parliament on 16 June 2026.

Nothing else in the timetable was touched. That is the part the headlines compressed away.

02Which obligations are already binding?

Application dates read from the operative text of Regulation (EU) 2024/1689 and its 2026 amendment. Dates verified against primary and contemporaneous sources on 12 September 2026.
ObligationApplies fromStatus today
Prohibited practices, Article 52 Feb 2025In force
AI literacy, Article 42 Feb 2025In force
General purpose AI duties, Chapter V2 Aug 2025In force
Transparency duties, Article 502 Aug 2026In force
Article 50(2) marking, legacy systems2 Dec 2026Pending
New Article 5 prohibition, synthetic intimate imagery2 Dec 2026Pending
High-risk, Annex III2 Dec 2027Deferred
High-risk, Annex I2 Aug 2028Deferred

Four of those eight rows are already live. One of them, Article 50, reaches almost every organisation shipping a chatbot or generating content, which is a far wider population than the high-risk one.

03Does the delay change what you should be doing this quarter?

It changes the deadline. It does not change the lead time.

The high-risk obligations are not a form to complete. They are technical documentation under Annex IV, a risk management system, data governance evidence, automatic logging, human oversight design and post-market monitoring. Organisations that have assembled that evidence pack before, for a medical device or a safety component, know it is measured in quarters rather than weeks.

Sixteen months is not generous against that. It is roughly the time it takes to get logging in place, run a full cycle of post-market monitoring, and have something an auditor can read.

The realistic use of the deferral is to reorder, not to stop. Classification work, which decides whether you are in scope at all, has no reason to wait, because the answer determines the size of everything downstream.

04What is the new prohibition arriving on 2 December 2026?

A prohibition, not a high-risk duty, which is why it survived the deferral.

The amendment adds to Article 5 a ban on AI systems generating non-consensual intimate imagery and child sexual abuse material. It carries its own compliance date of 2 December 2026 and applies regardless of whether an organisation operates any high-risk system.

Prohibitions sit at the top of the penalty scale and are not something a risk register manages down. For anyone operating an image or video generation feature, the relevant work is input and output filtering plus a documented account of what the filter is meant to stop, and it needs to exist before the date rather than after the first complaint.

05How should a programme plan be re-sequenced?

Sequencing prepared by The Guardrail from the application dates above. Editorial judgement, not a legal opinion, and not a substitute for advice on your own classification.
WorkstreamHard dateWhy it cannot wait for 2027
Article 50 disclosure and marking2 Dec 2026Already in force; only legacy marking has an extension
Prohibited practice review2 Dec 2026New Article 5 ban applies to everyone
High-risk classificationNo dateDetermines the scope of every later workstream
Annex IV technical documentation2 Dec 2027Depends on logging that must run for months first

The one sequencing error worth avoiding is treating classification as part of the deferred package. It is the input to the deferred package. An organisation that discovers in mid 2027 that three of its systems fall under Annex III has bought itself six months, not sixteen.

06Frequently asked questions

Did the EU delay the whole AI Act?

No. The amendment deferred the high-risk obligations only. Prohibited practices, AI literacy, general purpose AI duties and the Article 50 transparency rules were untouched and continue to apply on their original dates.

What are the new high-risk dates?

2 December 2027 for standalone Annex III systems and 2 August 2028 for AI embedded in products already regulated under Annex I. The previous dates were 2 August 2026 and 2 August 2027.

Should a high-risk programme slow down because of the deferral?

Not the evidence work. The deferral moves the date on which conformity must be demonstrable, not the difficulty of demonstrating it, and the technical documentation, logging and post-market monitoring take longer to assemble than the 16 months bought.

What happens on 2 December 2026?

Two things. Generative AI systems placed on the market before 2 August 2026 must meet the Article 50(2) machine-readable marking requirement, and the new prohibition on AI generated intimate imagery and child sexual abuse material starts to apply.

07References and method

  1. Regulation (EU) 2024/1689 as amended. The amendments were agreed provisionally on 7 May 2026 and approved by the European Parliament on 16 June 2026.
  2. Contemporaneous analyses of the agreement from Gibson Dunn, Morgan Lewis and Covington, consulted on 12 September 2026 for the deferral dates and the new Article 5 prohibition. Where they agree with each other and with the amending text we cite the date; we have not seen a consolidated official text.
  3. Regulation (EU) 2024/1689, Article 50, for the transparency duties, and Article 99 for the penalty ceilings of EUR 15 million or 3% of worldwide annual turnover.
  4. European Commission, Shaping Europe's digital future, transparency obligations factpage, for the 2 August 2026 application date of Article 50.
ML

, Regulatory Editor

Edits The Guardrail and writes its regulation coverage: what the text of a law actually requires, and on what date. Reach them at marta@theguardrailreport.com.