Change history

A full re-reading of every article the product cites against the amended text. Three claims overstated what the law asks of a small company; all three are corrected.

• Penalties for prohibited practices were quoted as EUR 35 million or 7% of worldwide turnover. That is the headline figure. Article 99 caps fines for SMEs at whichever of the two amounts is lower, and Regulation (EU) 2026/1744 extended the same cap to small mid-caps, so for a company of the size this product is written for the real exposure is a share of its own turnover. The figure is still quoted, now with the cap that applies to you.
• The Annex I question treated any AI inside a regulated product as a safety component. Article 6(1a), inserted by the Omnibus, excludes AI used solely for quality control, performance optimisation, automation, service efficiency or convenience, and the definition in Article 3(14) was narrowed to components whose purpose is to prevent or mitigate risks to health and safety. Industrial quality-control AI was being classified as high-risk when the amended text says plainly that it is not.
• Neither high-risk finding mentioned Article 111, which leaves systems already placed on the market before the regime applies outside it unless their design changes significantly afterwards. Public-authority systems are excepted and remain due by 2 August 2030. Both findings now open by asking when the system went into service, because that question can remove the entire obligation.
• Checked and found already correct: the Article 50 transparency duties and their 2 December 2026 content-labelling transition, which matches the new Article 111(4); the deferral dates for both Annex tracks; and the Article 5 prohibitions as amended.

Correction: Article 4 was described using the wording the Digital Omnibus repealed, and the new Article 5 prohibitions were missing.

• Article 4 was stated as an obligation to ensure a sufficient level of AI literacy. That was the wording before 27 July 2026, when Regulation (EU) 2026/1744 replaced the article outright. The current obligation is to take measures supporting the development of AI literacy, and it says expressly that no specific level need be guaranteed for any individual. Every document, finding and page describing it has been rewritten.
• The correction makes the obligation easier to meet rather than harder, so anyone holding the earlier edition was told to do more than the law asks, not less. What it does not change is that Article 4 has applied since 2 February 2025 and that national market surveillance authorities have enforced it since August 2026.
• The AI Usage Policy now also lists the prohibition on using AI to generate or manipulate non-consensual intimate imagery or child sexual abuse material, which Regulation (EU) 2026/1744 added to Article 5 and which applies from 2 December 2026. It is written as an internal rule effective on adoption rather than on that date.

First published edition, written against the AI Act as amended by the Digital Omnibus.

• The documents reflect the deferrals made by the Digital Omnibus in July 2026: Annex III high-risk obligations moved to 2 December 2027 and Annex I to 2 August 2028.
• They treat the deferral as a change of date and not of substance, so the register and policy still ask for the evidence the high-risk regime will require, on the basis that organisations which start late find the records were never kept.
• The obligations that are already in force are drafted as current rather than upcoming: AI literacy under Article 4 since 2 February 2025, and the Article 50 transparency duties from 2 August 2026.