Tuesday 17 November 2026
Business

Europe gives firms more time on AI rules: what changed

The EU has pushed back the strictest parts of its AI Act. Businesses get more time, but the obligations have not gone away.

Also in: FrançaisDeutschEspañol日本語

The Berlaymont building, headquarters of the European Commission in Brussels, with EU flags
Photo: KenSwe72, Wikimedia Commons (CC BY 4.0)

European companies that use artificial intelligence in sensitive areas have been given more time to prepare for the toughest AI rules. Under a package of changes known as the Digital Omnibus, the European Union has delayed the strictest obligations of its AI Act, the world's first broad law on artificial intelligence. Negotiators reached a deal in May 2026, and the amendment was published in the EU's Official Journal in July.

The new dates

The AI Act sorts AI systems by risk. The heaviest rules apply to "high-risk" systems, such as those used to screen job applicants, score people for credit, or decide access to education. Those obligations were due to apply from August 2, 2026. They now apply from December 2, 2027.

A second group, AI built into products that are already regulated, such as medical devices and machinery, now has until August 2, 2028.

What did not change

The delay is about timing, not substance. The requirements for high-risk systems, including risk management, data quality, human oversight and record-keeping, remain in the law. Companies that stop preparing now may find themselves short of time in 2027.

Other parts of the Act were not postponed. Bans on certain practices, such as social scoring and some uses of emotion recognition at work, have applied since February 2025. Rules for providers of general-purpose AI models, the large systems behind many chatbots, have applied since August 2025. The package also added new prohibited uses, including AI that creates non-consensual intimate images. Transparency duties, such as telling people when they are interacting with a chatbot and labelling some AI-generated content, also remain, with details and grace periods that companies should check against the official text.

Why the delay

Businesses and several member states had argued that the technical standards and guidance needed to comply were not ready in time. Without them, companies faced legal obligations they could not clearly meet. The Omnibus is part of a wider EU effort to simplify rules and support European competitiveness, a theme that has gained weight as the bloc tries to keep pace with the United States and China.

What companies should do now

  • Make a list. Record every AI system you build, buy or use, and what it is used for.
  • Sort by risk. Identify anything that could fall into the high-risk categories, such as hiring tools or credit scoring.
  • Talk to suppliers. If you buy AI from a vendor, ask how they plan to comply and what documents they will provide.
  • Use the extra time. Treat 2027 as a firm deadline and plan the work backwards from it.

The rules apply to companies outside Europe too, whenever their AI is used in the EU. For smaller firms just starting with AI, our practical guide for small businesses is a good place to begin.

What it means outside Europe

The AI Act has influence far beyond the EU, much like Europe's data protection law did before it. Companies in the Gulf, Asia and Africa that sell software or services to European customers must follow it, and many find it simpler to apply the same standards everywhere. Governments elsewhere, including in the Gulf, are writing their own AI policies and watching how Brussels handles enforcement. A delay in Europe gives everyone a little more time, but the direction of travel is the same: more documentation, more testing and more human oversight for AI that affects people's lives. For how banks are preparing for another technology shift, read our article on quantum-safe encryption.

Companies that use the extra time well will be in a stronger position, not only with regulators but with customers, who increasingly ask how AI decisions about them are made and checked.

Sources: Gibson Dunn; Future of Privacy Forum; AI Act explorer.

#Artificial Intelligence#Europe#Compliance#EU AI Act#Regulation

AJMN Newsroom · Editorial team

Share

Newsletter

The best of AJMN in your inbox, once a week.