Compliance Colleague · Answers · EU AI Act

When do the EU AI Act's high-risk obligations apply to Annex III and Annex I systems?

Standard: EU AI Act · Answered by Compliance Colleague on September 30, 2026 · Reviewed by Edward Casteloes CISSP, CISA

Question as asked: When do the EU AI Act high-risk obligations apply to Annex III systems such as CV-screening tools, and to Annex I product safety components?

Application dates for high-risk obligations

Annex III systems (e.g., CV-screening tools under Annex III point 4(a) – "analysing and filtering job applications, evaluating candidates"): The high-risk obligations in Chapter III, Sections 1–3 of Regulation (EU) 2024/1689 apply from 2 December 2027 (Art. 113, third paragraph, point (c)(i), as amended by Regulation (EU) 2026/1744). This date was pushed back by the Digital Omnibus from the original 2 August 2026 cut-off.

Annex I systems (safety components of products subject to third-party conformity assessment, e.g., machinery, medical devices): These obligations apply from 2 August 2028 (Art. 113, third paragraph, point (c)(ii), as amended). This too was deferred by the Omnibus from the original date.

Transitional rule for systems already on the market — Art. 111(2)

For high-risk systems (other than Annex X large-scale IT components) already placed on the market or put into service before the relevant application date above, Regulation (EU) 2024/1689 applies to providers and deployers only if the system undergoes a significant design change on or after that date (Art. 111(2), as replaced by Regulation (EU) 2026/1744). Key points:

  • This is a grace period, not an exemption. A legacy CV-screening tool placed on the market before 2 December 2027 stays outside Chapter III obligations until it undergoes a significant design change on or after that date — at which point full compliance, including conformity assessment, is triggered.
  • A significant change made before the application date does not start the obligations early; no one is subject to Chapter III high-risk obligations before 2 December 2027 (Annex III) or 2 August 2028 (Annex I), regardless of updates made in the interim.
  • The grace period is without prejudice to Art. 5 (prohibited practices) and does not affect Art. 50 transparency duties, which run on their own separate timeline.
  • Public-authority systems are an exception: high-risk systems intended for use by public authorities must comply by 2 August 2030 regardless of whether a significant design change occurs (Art. 111(2), second sentence).
  • A system first placed on the market after the applicable date gets no grace period at all.

Missing fact that would change the answer: whether the CV-screening tool (or Annex I component) was already placed on the market/put into service before its respective application date, and whether the deployer is a public authority. If it's a new system launching after 2 December 2027 (Annex III) or 2 August 2028 (Annex I), full obligations apply immediately with no transition.

AI-generated research aid, not legal or audit advice. This answer reflects the standard as of the date above; verify it against the official text before relying on it.

How Compliance Colleague made this

  1. The question above was entered in Compliance Colleague with EU AI Act selected as the standard.
  2. The answer came back as shown, grounded in the standard’s own text, with the article or clause cited behind each point. Nothing was added by hand.
  3. Edward Casteloes CISSP, CISA checked the citations against the official text before it was published here.

Check your own policy against EU AI Act →

This is one answer from Compliance Colleague, which answers questions like this against 30+ security, privacy and AI governance standards, with the clause cited, and reviews your policies against them. Ask your own question free, no card required →

More answers

Related guides