European Parliament votes to simplify AI Act compliance rules

The European Parliament has adopted a package of amendments simplifying the implementation of the EU’s harmonised rules on artificial intelligence, in a first-reading vote under the ordinary legislative procedure. The amendments, published in the Official Journal of the European Union (C series, C/2026/4045, of 2 September 2026), were approved on 26 March 2026 and target the “Digital Omnibus on AI”, a proposal amending Regulation (EU) 2024/1689 (the AI Act) and Regulation (EU) 2018/1139.

The text, based on the Commission proposal COM(2025)0836 (2025/0359(COD)), was referred back for interinstitutional negotiations to the responsible committees under Rule 60(4), fourth subparagraph, of Parliament’s Rules of Procedure (report A10-0073/2026). Its aim is to make the AI Act easier to apply for businesses of all sizes while preserving the level of protection the regulation guarantees.

Lighter obligations for SMEs and small mid-caps

A central theme of the amendments is proportionality for smaller players. The Parliament notes that 99.8% of all Union companies are small and medium-sized enterprises, most of them micro and small, and introduces formal definitions for SMEs and for “small mid-cap enterprises” (SMCs) aligned with Commission Recommendations 2003/361/EC and (EU) 2025/1099.

• The AI literacy obligation in Article 4 is softened from a stringent horizontal duty to a requirement to “support” staff literacy, with the European Commission tasked to issue practical guidance.
• Registration of certain AI systems in the EU database under Article 6(3) is simplified by streamlining the content required in Section B of Annex VIII.
• SMEs, including startups, and SMCs are granted priority access to AI regulatory sandboxes, including those established at Union level by the AI Office.

New safeguards and clearer boundaries

The amendments also tighten protection in specific areas. A new recital calls for an explicit prohibition on AI systems that alter, manipulate or artificially generate realistic images or videos depicting sexually explicit activities or intimate parts of an identifiable person without consent — the so-called “nudification” applications — while preserving technical development capabilities.

• The legal basis for processing special categories of personal data for bias detection and correction is extended to providers and deployers of other AI systems and models, under the same conditions and safeguards as the existing Article 10(5).
• The AI Office is to be given adequate human, financial and technical resources to exercise its new powers.
• Alignment is sought with Regulation (EU) 2024/2847 on cybersecurity, so that high-risk AI systems meeting its essential cybersecurity requirements are deemed to comply with Article 15 of the AI Act.

Deadlines and next steps

The amendments adjust transitional deadlines for providers of AI systems generating synthetic audio, image, video or text content placed on the market before 2 August 2026, who must comply with Article 50(2) by 2 November 2026. The matter now returns to interinstitutional negotiations between Parliament, the Council and the Commission before a final text can be adopted.

For the thousands of companies and developers of AI systems in the EU, the package promises a lighter compliance burden and clearer guidance, while keeping the core safeguards of the AI Act intact. Its outcome will shape how innovation and regulation are balanced across the single market in the coming years.

Source: Official Journal of the European Union, C series, C/2026/4045, 2 September 2026 (official reference: C/2026/4045, ELI http://data.europa.eu/eli/C/2026/4045/oj).