Trade War

EU delays high-risk AI duties while December content rules approach

The enacted AI Omnibus gives high-risk systems more time, but existing prohibitions, general-purpose model rules and approaching content-marking duties require separate attention.

The law 17 November 2026

The European Union has postponed its main high-risk artificial-intelligence requirements, while retaining a nearer deadline for some synthetic-content providers. The AI Omnibus entered into force on 27 July 2026. Most listed high-risk uses now face the main requirements from 2 December 2027, with covered product-related systems following on 2 August 2028, according to the European Parliament's legislative record.1

The change matters to companies selling software, machinery and digital services into Europe because the relevant timetable depends on the system's purpose. The Commission's explanation distinguishes high-risk applications, transparency duties and general-purpose models. Moving one deadline does not make those categories interchangeable.2

The product and its use determine the obligations

The Commission identifies recruitment software, certain credit assessments, educational decisions and critical-infrastructure safety components among the uses potentially classified as high-risk. Their requirements include documented risk management, suitable training data, activity records, information for users, human oversight and adequate accuracy and cybersecurity. These are conditions governing how a system reaches and operates in the market.2

Other obligations have already started. The Commission dates the original prohibited-practice rules to February 2025 and the general-purpose model rules to August 2025. Its guidance describes separate transparency and copyright responsibilities for model providers, with additional risk assessment and mitigation for models presenting systemic risks. A supplier therefore needs to establish which role it performs as well as what software it sells.2

The Council describes the revised approach as seeking:

une mise en œuvre plus harmonisée

More harmonised implementation.3

That is the Council's stated objective. Its adoption announcement also describes changes to supervision and to the relationship between the AI Act and product legislation. The intention is to reduce overlapping procedures while keeping relevant safeguards, rather than to create an unrestricted route for technology imports.3

Machinery suppliers welcome a change that medical technology did not receive

Industry association DIGITALEUROPE welcomed the machinery provisions when negotiators reached their agreement in May. It argued that keeping machinery under a single sectoral compliance route could avoid duplicate assessments for manufacturers. The association nevertheless criticised the treatment of medical technology, saying those suppliers would still face overlapping regulatory requirements. Those are the industry's assessments of the agreement, rather than measured savings already achieved.4

DIGITALEUROPE also called for industry participation in the technical work that would put the machinery changes into practice. Its reaction illustrates why a common headline deadline cannot describe every exporter's position: the sectoral legislation and the eventual standards remain commercially important. The association's criticism concerns the way separate legal requirements interact, not an exemption for all AI embedded in equipment.4

The Parliament's record confirms that the machinery legislation moves to a different part of the AI Act's product framework, with relevant requirements to be integrated through sectoral measures. For suppliers, the unresolved implementation work concerns the detailed requirements applying to their particular product.1

The nearer dated event is 2 December 2026. Providers of synthetic-content systems already marketed before 2 August 2026 must meet the relevant content-marking obligations by then; the new prohibition concerning specified non-consensual intimate and child sexual-abuse material also starts that day.1

Sources

  1. European Parliament. Artificial intelligence: simplification of certain rules, final act summary. 24 July 2026European Union's camp · legislative record
  2. European Commission. Législation sur l’IA [AI legislation]. Accessed 17 November 2026European Union's camp · government · in French
  3. Council of the European Union. Intelligence artificielle: le Conseil donne son feu vert définitif à la simplification et à la rationalisation des règles [Artificial intelligence: Council gives final green light to simplify and streamline rules]. 29 June 2026European Union's camp · government · in French
  4. DIGITALEUROPE. AI Omnibus: DIGITALEUROPE welcomes machinery breakthrough but warns medtech left behind. 7 May 2026Industry camp · trade association

All articles