Brussels proposed easing the AI Act for medical devices. Lawmakers kept it in place.

The European Commission announced on 16 December a package of measures to modernise the EU’s health sector, proposing to move most medical devices into a lighter category of the AI Act so that a diagnostic algorithm or a surgical planning tool would fall within the law’s scope only to a limited extent. The AI Act’s medical device provisions sit among the most demanding in the law, because an algorithm that misreads a scan carries a different order of risk than one that mistargets an advert.

Two days later, on 18 December, the United Kingdom’s Medicines and Healthcare products Regulatory Agency opened a call for evidence on how AI in healthcare should be regulated, asking clinicians, industry and the public what rules they would want before the UK commits to any of its own. Where Brussels was already rewriting its statute book, London was still asking the question.

That contrast did not hold. In the negotiations that followed, known as the AI Omnibus, the Council of the EU and the European Parliament rejected the Commission’s central idea for health technology: diagnostic devices and in vitro diagnostics, the laboratory tests covered by a parallel law known as the IVDR, will stay fully within the AI Act’s strictest, high-risk tier, with compliance deadlines pushed back to August 2028 rather than the broad exemption the Commission had proposed.

The proposed broad inapplicability of the AI Act to medical devices and IVDs will therefore not be implemented.

Lawmakers chose to keep the safety architecture intact and buy manufacturers time instead, a more cautious outcome than the Commission’s own proposal had envisaged.