On 16 June 2026, the European Parliament gave its final approval of the targeted amendments to the EU Artificial Intelligence Act (“AI Act”) as part of the AI Digital Omnibus Regulation (2025/0359/EU) (the “Proposal”). The vote passed with 423 votes in favour, 57 against and 174 abstentions. The Proposal aims to simplify and address some of implementation challenges of the AI Act. The Parliament’s adopted text is generally in line with the provisional agreement reached by the Parliament and Council on 7 May 2026 (“the provisional agreement”) in regard to the Proposal (previously discussed here).
This article sets out a high-level summary of the key amendments approved by the Parliament.
Key amendments
- Extension of implementation timeline for high-risk AI systems: The adopted text confirms the postponement of the implementation timeline for the rules on high-risk AI systems. These rules will apply from:
- 2 December 2027 for standalone high-risk AI systems under Article 6(2) and Annex III of the AI Act (eg, AI systems involving biometrics, critical infrastructure, education, employment, essential services, law enforcement, justice and border management), and
- 2 August 2028 for AI systems embedded as safety components and covered by EU sectoral legislation under Article 6(1) and Annex I of the AI Act.
- Transparency and watermarking obligations: In line with the provisional agreement, Parliament has approved the extension of the implementation timeline for the transparency and watermarking obligations under Article 50(2) AI Act until 2 December 2026. By this time, AI-generated content will have to be labelled in a machine-readable way to increase transparency. All other transparency obligations will continue to apply from 2 August 2026.
- Prohibition on non-consensual intimate images and child sexual abuse material: Another priority amendment confirmed by Parliament is the ban on “nudifier” apps. Parliament’s adopted text prohibits AI systems that generate non-consensual intimate images and child sexual abuse material. This applies to providers, as well as deployers using the AI systems for such purposes. Companies will have until 2 December 2026 to bring their systems in compliance. The Parliament’s text also clarifies, in line with the provisional agreement, that this ban would apply where (i) such generation is the intended purpose of the AI system and (ii) where the providers and deployers have not put in place adequate technical safeguards to prevent the generation of such content.
- Machinery products and sectoral legislation: A sticking point of the trilogue negotiations, the adopted text follows the provisional agreement, by carving out machinery regulation (Regulation (EU) 2023/1230) from direct applicability of the AI Act (ie, moving it from Annex I Section A to Section B). This means machinery products will only need to comply with requirements of EU sectoral legislation, while being subject to limited obligations under the AI Act to ensure an equivalent level of health and safety.
- Clearer definition of “safety component”: Under the original text of Article 6(1) of the AI Act, an AI system is classified as high-risk where it is intended to be used as a safety component of a product, or the AI system is itself a product, covered by the EU harmonisation legislation listed in Annex I and is required to undergo conformity assessment. In this regard, Parliament’s adopted text amends the definition of a “safety component” in Article 3(14) to clarify that “a component fulfils a safety function where its intended purpose is to prevent or mitigate risks to health and safety of persons or property”. As such, the amendment seeks to exclude from the high-risk category products with AI functions that merely assist users or optimise performance, if their malfunction or failure does not create a health and safety risk.
- Processing of special categories of personal data for bias detection and correction: In line with the provisional agreement, Parliament has confirmed that processing of special categories of personal data to detect and correct biases can only be carried out in certain exceptional circumstances and subject to adequate safeguards. In this regard, the Parliament reinstates the standard of strict necessity for providers and deployers of high-risk and non-high-risk AI systems and models.
- Small mid-cap enterprises: Again in line with the provisional agreement, Parliament’s adopted text extends privileges afforded to small and medium-sized enterprises in Article 99 of the AI Act to small mid-cap enterprises in order to support their growth.
- Stronger EU AI Office enforcement: The adopted text also clarifies the scope of responsibilities and powers of the EU AI Office. In particular, the adopted text grants AI Office exclusive competence to supervise certain AI systems integrating general-purpose AI models (such as where the model and system are developed by the same provider) and those embedded into very large online platforms and very large search engines (as defined in the Digital Services Act). Notwithstanding this, the adopted text facilitates exceptions where national authorities will remain competent.
- AI literacy obligation softened: The adopted text requires providers and developers to take measures to “support the development of AI literacy”. This replaces the current Article 4 AI Act text which requires providers and deployers to “ensure…a sufficient a level of AI literacy”. The text also requires the Commission and the Member States to support and facilitate the efforts of providers and deployers in this regard, and for the Commission to publish practical examples of how to comply with this obligation.
- EU database registration requirements: Finally, and in line with the provisional agreement, Parliament has reinstated the obligation in the AI Act (previously removed in the Proposal) for providers to register certain high-risk AI systems in the EU database. This obligation applies even where providers have concluded that such systems are not high-risk in accordance with Article 6(3) of the AI Act (ie, where they do not pose a significant risk of harm to individuals’ health, safety or fundamental rights). However, this requirement will be simplified by removing some of the required content that must be provided under Section B of Annex VIII of the AI Act (namely, providing the status of the AI system and a scanned copy of a certificate issued by a notified body, where applicable), thereby aiming to strike a better balance without undermining the intended protection.
Next Steps
Parliament’s text must now be formally adopted by the Council and then be submitted for a legal / linguistic review before it can enter into force. This process is expected to be completed prior to 2 August 2026 (the current start date for the rules on high-risk AI systems under the AI Act).
Despite the delayed implementation of the high-risk AI rules, it is critical that businesses continue with their AI Act compliance efforts, and finalise their AI governance frameworks as soon as possible.
Contact Us
For more information or assistance with AI compliance, please contact any member of our Technology and Innovation Group or your usual Matheson contact.
