EU AI Act High-Risk Deadlines Moved to December 2027 and August 2028

Regulation (EU) 2024/1689, the Artificial Intelligence Act, is the EU’s horizontal law on artificial intelligence. It was adopted on 13 June 2024, published in the Official Journal on 12 July 2024 and entered into force on 1 August 2024. Its obligations apply in stages under Article 113, rather than all at once.

The first obligations applied from 2 February 2025: the prohibitions on certain AI practices in Article 5 and the AI literacy provision in Article 4, which originally required providers and deployers to ensure a sufficient level of AI literacy among staff. Obligations for providers of general-purpose AI models, together with governance structures, applied from 2 August 2025; providers of models placed on the market before that date have until 2 August 2027 to comply.

For research and health, the Act matters in two ways. First, tools such as large language models used for drafting, translation or literature screening are general-purpose AI systems whose providers carry obligations, while organisations that use them are deployers. Second, AI that is a safety component of, or itself is, a product covered by EU product legislation listed in Annex I, which includes medical devices, can be classified as high-risk.

The European Commission proposed a Digital Omnibus on AI in November 2025 to simplify and delay parts of the Act. A provisional political agreement was reached on 7 May 2026, the European Parliament endorsed it on 16 June 2026 and the Council adopted it on 29 June 2026, according to law firm summaries. The amending regulation, reported as Regulation (EU) 2026/1744, was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026; EUR-Lex carries a consolidated text of the AI Act dated 27 July 2026.

The omnibus postpones the high-risk obligations. Stand-alone high-risk systems listed in Annex III now face obligations from 2 December 2027. High-risk AI embedded in products regulated under Annex I legislation, including medical devices, moves from 2 August 2027 to 2 August 2028. Medical devices and in vitro diagnostics did not receive a full carve-out, so they remain subject to both the AI Act and the medical device regulations, with a mechanism to limit AI Act requirements where those regulations already cover similar ground, to be set out in implementing acts and Commission guidance.

Commentators report that the omnibus also softened the AI literacy duty, so that Article 4 now asks providers and deployers to take measures to promote AI literacy rather than to ensure a sufficient level. Sources differ on the exact wording, so organisations should check the consolidated text.

One summary also reports that the omnibus extends simplifications available to small and medium-sized enterprises to small mid-cap companies, which may be relevant to smaller sponsors, CROs and health technology developers.

The transparency obligations in Article 50, which include marking AI-generated content, apply from 2 August 2026 and were not postponed. Summaries report a transitional period until 2 December 2026 for systems already on the market to implement machine-readable marking of AI-generated content. General-purpose AI model obligations were also not delayed.

Article 2(6) excludes AI systems and models specifically developed and put into service for the sole purpose of scientific research and development. Commentary describes this as covering AI as an object of study rather than AI as a tool of study. A university that builds an experimental model for a research project may benefit from the exclusion, but once the model is deployed operationally, or where researchers use commercial AI tools for writing, analysis or administration, the institution acts as a deployer and the relevant obligations apply.

Organisations should inventory the AI tools they use, identify whether they act as provider or deployer for each, and confirm that no use falls within the prohibited practices. Training programmes developed for the AI literacy provision remain good practice even after the wording change. Sponsors developing AI-enabled medical devices or software should plan conformity assessment around the August 2028 date, while monitoring implementing acts on overlap with the medical device rules.

Medical writers, publishers and communications teams that generate text, images or other content with AI should review how the Article 50 transparency duties apply to their outputs and to the tools they use. This brief summarises public sources for information and is not legal advice; organisations should consult the consolidated text on EUR-Lex.