Blogs / 

EU Reaches Political Agreement on Amendments to the AI Act

IT, Privacy & Cybersecurity

8 May 2026

Written by

Menno de Wijs

Blog Image

The European Union has reached a political agreement on an amendment to the AI Act. This does not yet constitute final legislation, but the direction is clear: the introduction of various obligations for high-risk AI systems is likely to be postponed. The amendment follows widespread concerns among businesses, regulators and Member States regarding the feasibility of implementing the rules, partly because sufficient standards and practical guidance are still lacking.

What Has Been Agreed?

According to the provisional agreement, the main deadlines for high-risk AI systems will be postponed. For AI systems listed in Annex III, the deadline is expected to move from 2 August 2026 to 2 December 2027. For AI systems listed in Annex I, such as systems integrated into regulated products, the deadline will be extended to 2 August 2028.

In addition, the political agreement includes a ban on AI systems that generate non-consensual sexual or intimate content, a postponement of certain watermarking and detection obligations for generative AI, and broader possibilities for processing sensitive personal data for the purposes of bias detection and correction.

No Final Legislation Yet

It is important to emphasise that this concerns a “political agreement” between the European Parliament and the Council of the EU. Formal adoption still has to take place, and from a legal perspective the current provisions of the AI Act therefore remain in force. However, in European legislative practice, such an agreement generally means that the main elements are unlikely to change fundamentally.

No Reason to Sit Back

Although the possible postponement of deadlines will be welcomed by many organisations, it does little to change the core of the AI Act. The obligations regarding governance, documentation requirements and responsibilities surrounding AI remain in place. For organisations using AI in areas such as HR, procurement, healthcare, education or public services, preparing in advance remains advisable. This agreement should primarily be viewed as a measure to ease implementation pressure.

Questions?

Do you have any questions? Please contact Menno de Wijs or one of our other specialists within the IT, Privacy & Cybersecurity Team.

Newsletter

Would you like to receive a monthly overview of updates and blogs in your inbox? Subscribe to our newsletter!