Analysis
Articles
Shorter, dated pieces on one question at a time: what an amendment actually changed, how a test is applied in practice, where a determination is contested. Each one links back to the guide that covers the underlying rule.
Last reviewed 4 August 2026Regulation (EU) 2024/1689, as amended by the Digital Omnibus
6 August 2026Which conformity assessment route applies to a high-risk AI system?Article 43 sets three routes, not two: Annex VI internal control, an Annex VII notified body, or the Annex I sectoral procedure.6 August 2026Who has to do a fundamental rights impact assessmentThe Article 27 FRIA is a deployer duty, not a provider one. Who it catches, its six required elements, and why it starts on 2 December 2027.6 August 2026How serious incident reporting works under Article 73Article 73 makes providers of high-risk AI systems report serious incidents. The four limbs of the definition, the 15, 10 and two-day clocks, and who files.5 August 2026When is emotion recognition banned, and when is a notice enough?Article 50(3) has applied since 2 August 2026 and binds the deployer. In the workplace and in education, Article 5(1)(f) prohibits the same system.5 August 2026Does fine-tuning make you the provider?Article 25 makes you the provider of a high-risk AI system on three triggers, and only one of the three involves changing anything technical.5 August 2026What does Article 5 of the AI Act prohibit?Article 5 has banned eight AI practices since 2 February 2025 and two more from 2 December 2026, and no conformity assessment makes any of them lawful.4 August 2026What an AI chatbot disclosure has to sayArticle 50(1) applied on 2 August 2026. What a compliant AI disclosure says, where it goes on six surfaces, and the patterns that fail 50(5).4 August 2026Does the 2 December 2026 marking grace period cover you?The 50(2) marking grace covers only generative systems placed on the EU market before 2 August 2026. How to tell which side of that line you sit on.3 August 2026Meeting the Article 50(2) marking dutyArticle 50(2) wants marking that is effective, interoperable, robust and reliable. What that means in practice, and which systems the December date helps.31 July 2026What to ask an AI vendor before you signTwelve questions that establish whether a vendor is the provider, what documentation exists, and which obligations land on you the moment you sign.30 July 2026Labelling deepfakes under Article 50(4)Article 50(4) puts the deepfake labelling duty on the deployer, not the tool vendor, and the artistic carve-out is narrower than agencies assume.29 July 2026The open-source exemptions, and what they do not coverFree and open-source AI gets two narrow exemptions under Articles 2 and 53. Neither survives high-risk classification, Article 50, or systemic-risk scale.28 July 2026The Article 4 AI literacy dutyArticle 4 has bound providers and deployers since 2 February 2025, and it reaches every member of staff who operates an AI system on your behalf.27 July 2026Where the EU AI Act and the GDPR overlapThe AI Act and the GDPR stack rather than replace one another. Where the duties overlap, where they diverge, and which regulator enforces which breach.25 July 2026What is high-risk AI?High-risk is a classification, not a severity rating. Where it sits among the AI Act's risk categories, what it catches and what it does not.