On 16 June 2026, the European Parliament approved amendments to the AI Act as part of the Digital Omnibus package. The changes include the postponement of key obligations for high-risk AI systems, a strengthened prohibition on AI systems used to generate non-consensual intimate content and child sexual abuse material (CSAM), adjustments to rules on transparency, bias detection, regulatory sandboxes and AI literacy, and enhanced powers for the European AI Office. However, 2 August 2026 remains a key compliance date, as the transparency obligations under Article 50 remain largely unchanged. The amendments must still be formally adopted by the Council of the European Union and published in the Official Journal before becoming legally binding.
On 16 June 2026, the European Parliament approved, in a final vote, amendments to certain rules of the European Union Artificial Intelligence Act (AI Act) as part of the Digital Omnibus package.
The text was approved with 423 votes in favour, 57 against and 174 abstentions.
Key amendments approved:
- Postponement of deadlines for high-risk AI systems – the dates from which the obligations applicable to high-risk Artificial Intelligence (AI) systems will apply have been postponed. In particular: (i) stand-alone high-risk AI systems, including, among others, those used in recruitment, credit scoring, law enforcement, education and border control, will be subject to the relevant obligations from 2 December 2027; and (ii) AI systems embedded as safety components in products regulated by sector-specific European Union (EU) legislation, such as medical devices, machinery and vehicles, will become subject to the relevant obligations from 2 August 2028.
- Prohibition of “nudifier” apps and CSAM – article 5 (prohibited AI practices) will prohibit AI systems that generate or manipulate non-consensual intimate images, videos or audio, or child sexual abuse material (CSAM). Providers will not be allowed to place such systems on the EU market where their purpose is to create such content, or where they do not include appropriate technical safeguards to prevent the creation of such material. The prohibition will also apply to deployers using those systems for such purposes. Companies will have until 2 December 2026 to bring their systems into compliance.
- Labelling of AI-generated content (watermarking) – AI systems placed on the market before 2 August 2026 will benefit from a four-month grace period, until 2 December 2026, before the watermarking obligation under Article 50(2) becomes applicable. This obligation requires providers to incorporate machine-readable markers into AI-generated or manipulated content.
- Products regulated by sector-specific legislation – the package introduces mechanisms to avoid the duplication of requirements arising from the AI Act and EU sectoral safety legislation, notably in the case of products such as medical devices, machinery and toys.
- Detection and correction of bias – the agreement extends the existing recognised legal basis for the processing of special categories of personal data for the detection and correction of bias (previously limited to providers of high-risk AI systems) to all AI systems and general-purpose AI models, subject to a strict necessity test.
- Regulatory sandboxes – the deadline for Member States to establish AI regulatory sandboxes is postponed to 2 August 2027, one year later than initially planned.
- European AI Office – the package strengthens the supervisory and enforcement role of the European AI Office. In particular, the Office will have exclusive competence over AI systems developed on the basis of general-purpose AI models where the model and th system are developed by the same provider.
- AI literacy obligation – the AI literacy obligation under Article 4 is reformulated and softened: providers and deployers will be required to support the development of AI literacy among their staff, rather than ensure a specific level of literacy.
What remains unchanged
Despite the amendments introduced by the Digital Omnibus package, 2 August 2026 remains a relevant date for compliance purposes. The transparency obligations under Article 50 remain largely unchanged, meaning that companies subject to those obligations should continue preparing for that date.
Next steps
Before entering into force, the text must still be formally adopted by the Council of the European Union and published in the Official Journal of the European Union. Accordingly, the new deadlines and amendments will only become legally binding after formal adoption and official publication.
The Technology team will continue to closely monitor legislative and regulatory developments applicable to the artificial intelligence sector and remains fully available to address any further questions.
