Statistics

Global AI in health regulation tracker

A living status table of the rules governing AI in healthcare — what is in force, what is proposed, and the exact effective dates — across the EU, the US FDA, the UK MHRA, the WHO, and other jurisdictions, each row tied to a primary regulator source. As of July 2026.

By Jonas WeirReviewed by Jonas Weir · editorial reviewUpdated

The short version

  • EU AI Act: the Article 4 AI-literacy duty for providers and deployers has applied since 2 February 2025, and the general regime — including stand-alone high-risk systems listed in Annex III — is set to apply from 2 August 2026 under the enacted Regulation.
  • A proposed Digital Omnibus on AI would defer stand-alone Annex III high-risk obligations to 2 December 2027. It was proposed on 19 November 2025 and approved by the European Parliament on 16 June 2026, but as of July 2026 it is not yet formally adopted — so the 2 August 2026 date still stands in law.
  • US FDA: the marketing-submission guidance on Predetermined Change Control Plans (PCCPs) is final; the broader total-product-lifecycle guidance remains a January 2025 draft.
  • UK MHRA runs the AI Airlock regulatory sandbox for AI as a medical device; the pilot and a second phase are complete (reports October 2025 and July 2026), with a third phase in design.
  • The WHO's January 2024 guidance on large multi-modal models is advisory; binding device rules sit with national regulators — Saudi Arabia's SFDA (MDS-G010), Australia's TGA, and others.

The rules that govern AI in healthcare are moving faster than almost any other part of the field, and they differ sharply by jurisdiction. This page tracks the current status of the major instruments — what is in force, what is only proposed, and the exact dates that matter — with every row tied to a primary regulator source. It is a status tracker for orientation and general information; confirm the live text with the regulator before acting. As of July 2026.

The status table at a glance

Jurisdiction / instrumentStatusKey dateSource
EU AI Act — Article 4 AI-literacy dutyIn forceApplies since 2 Feb 20251
EU AI Act — high-risk regime (Annex III, Art. 6(2))Enacted, awaiting applicationSet to apply 2 Aug 20261
EU Digital Omnibus — Annex III high-risk deferralProposed, not yet adoptedWould move to 2 Dec 20272
US FDA — PCCP marketing-submission guidanceFinalFed. Register Dec 20243
US FDA — lifecycle (TPLC) guidanceDraftIssued Jan 20254
UK MHRA — AI Airlock sandboxPilot + Phase 2 completeReports Oct 2025, Jul 20265
WHO — large multi-modal model guidanceAdvisoryPublished 18 Jan 20246
Saudi Arabia — SFDA MDS-G010Binding, in forceGuidance MDS-G010 (v. 2023)7
Australia — TGA AI consultationReport publishedJul 20258
FDA / Health Canada / MHRA — GMLP principlesPublishedOct 20219

Each row is unpacked below with its exact wording and effective date.

European Union: what is live, and what is in flux?

The EU AI Act (Regulation (EU) 2024/1689) applies in stages. The first stage that touches almost every organization using AI is already live: the Article 4 AI-literacy duty, which requires providers and deployers to ensure staff and others operating their systems have a sufficient level of AI literacy, has applied since 2 February 2025 1.

The stage most healthcare builders watch is the high-risk regime. Many clinical decision-support and diagnostic tools fall under Annex III, and under the enacted Regulation the obligations for stand-alone high-risk systems are set to apply from 2 August 2026 1.

Here is the nuance that matters most, and the reason this page carries a ninety-day cadence. On 19 November 2025 the European Commission proposed a Digital Omnibus on AI that would defer stand-alone Annex III high-risk obligations to 2 December 2027 (and to 2 August 2028 for high-risk AI embedded in regulated products). The proposal advanced quickly — the European Parliament approved the agreed text on 16 June 2026 — but as of July 2026 it has not been formally adopted and published in the Official Journal 2. Until it is, the enacted date stands: in law, the 2 August 2026 timeline still applies. Treat the deferral as likely but not yet certain, and watch for the Official Journal notice.

United States: where does FDA guidance stand?

The FDA regulates AI when it meets the device definition, and it has issued two central guidance documents.

The first is final: Marketing Submission Recommendations for a Predetermined Change Control Plan (PCCP) for Artificial Intelligence-Enabled Device Software Functions. Its underlying Federal Register notice dates to December 2024, and the current fda.gov guidance page carries an August 2025 date 3. A PCCP lets a manufacturer pre-specify certain model changes so it can update an AI-enabled device without filing a new marketing submission each time.

The second is still a draft: Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations, issued in January 2025 and, as of July 2026, not yet finalized 4. It is the more comprehensive of the two, covering design, validation, transparency, and post-market monitoring across the total product lifecycle (TPLC). Because it remains open, its recommendations can shift before they settle.

United Kingdom: a sandbox, not yet a statute

The UK's approach to AI as a medical device runs partly through the MHRA AI Airlock, a regulatory sandbox where developers and the regulator test how novel or higher-risk AI can operate safely within the existing device framework. The pilot and a second phase are both complete — the Pilot Programme Report was published in October 2025 and the Phase 2 Programme Report in July 2026 — and a third phase is in design 5. The Airlock is a learning mechanism that feeds future rules rather than a binding regime in itself, so read it as a signal of direction.

What role does the WHO play?

The World Health Organization issued Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models on 18 January 2024, with more than 40 recommendations for governments, developers, and providers 6. It carries real influence over national policy, but it is advisory: it sets no market-authorization requirement on its own.

Other jurisdictions in brief

Binding device rules live with national regulators, and several have moved early:

  • Saudi Arabia (SFDA): MDS-G010 sets binding requirements for manufacturers of AI/ML medical devices to obtain Medical Device Marketing Authorization — one of the region's earliest binding regimes, per a peer-reviewed analysis in Asian Bioethics Review (2024) of the 2023 guidance 7.
  • Australia (TGA): regulates AI under a technology-agnostic device framework; its consultation Clarifying and strengthening the regulation of Artificial Intelligence (AI) ran in late 2024, with an outcomes report published in July 2025 and further work approved 8.
  • International harmonization: Good Machine Learning Practice guiding principles were released in October 2021 jointly by the FDA, Health Canada, and the MHRA — an early tri-regulator step still cited across markets 9.

How to read these numbers

Four cautions travel with every row. First, status is perishable: a "draft" becomes "final," and a "proposed" date becomes law or lapses — the EU high-risk deadline is the clearest live example. Second, guidance is not legislation; an FDA guidance or a WHO recommendation shapes expectations but binds differently from a regulation or a statute. Third, scope varies: the same tool can be a regulated device in one market and outside the device definition in another. Fourth, effective dates are staggered within a single law — the EU AI Act alone has separate dates for literacy, general-purpose models, Annex III systems, and product-embedded systems.

We revisit this page every ninety days and whenever a tracked instrument changes status, scope, or date. For adjacent numbers, see our tracker of FDA-cleared AI devices by year and specialty and the AI scribe adoption statistics.

Sources and method. Every row is drawn from a primary regulator source — EUR-Lex for the EU AI Act, fda.gov for FDA guidance, gov.uk for the MHRA, who.int for the WHO, sfda.gov.sa for the SFDA, and the TGA consultation portal — listed and numbered above. Where a date reflects a publishing or notice event rather than a statutory effective date, we say so in the row. Dates and statuses are current as of July 2026.

Questions & answers

  • Is AI in healthcare regulated yet?

    It depends on the jurisdiction and the tool. AI that meets the definition of a medical device is already regulated in most major markets — by the FDA in the US, the MHRA in the UK, the SFDA in Saudi Arabia, and the TGA in Australia, among others. The EU AI Act adds a horizontal layer on top, with its AI-literacy duty already in force and its high-risk regime set to apply from 2 August 2026. General-purpose and advisory AI that falls outside the device definition is governed mostly by guidance rather than binding rules today.

  • When do the EU AI Act high-risk rules apply to health AI?

    Under the enacted Regulation, obligations for stand-alone high-risk systems listed in Annex III are set to apply from 2 August 2026. A proposed Digital Omnibus would push that to 2 December 2027, but as of July 2026 the Omnibus has not been formally adopted, so the 2 August 2026 date remains the one in law. This page is updated when that changes.

  • What is the difference between the FDA's PCCP and lifecycle guidance?

    The PCCP guidance — final — lets a manufacturer pre-specify certain changes to an AI-enabled device so it can update the model without a new marketing submission each time. The broader lifecycle guidance, still a January 2025 draft, covers design, validation, and post-market monitoring across the total product lifecycle. One is settled policy; the other is still open for comment.

Sources

  1. Regulation (EU) 2024/1689 of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Article 113. Official Journal of the European Union, 2024. eur-lex.europa.eu/eli/reg/2024/1689/oj
  2. European Parliament, Legislative Train Schedule: Digital Omnibus on AI. Accessed July 2026. www.europarl.europa.eu/legislative-train/package-digital-package/file-digital-omnibus-on-ai
  3. US Food and Drug Administration. Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions (Final Guidance). www.fda.gov/regulatory-information/search-fda-guidance-documents/marketing-submission-recommendations-predetermined-change-control-plan-artificial-intelligence
  4. US Food and Drug Administration. Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations (Draft Guidance, issued January 2025). www.fda.gov/regulatory-information/search-fda-guidance-documents/artificial-intelligence-enabled-device-software-functions-lifecycle-management-and-marketing
  5. UK Medicines and Healthcare products Regulatory Agency. AI Airlock: the regulatory sandbox for AIaMD (collection). Accessed July 2026. www.gov.uk/government/collections/ai-airlock-the-regulatory-sandbox-for-aiamd
  6. World Health Organization. Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models. 18 January 2024. www.who.int/publications/i/item/9789240084759
  7. Aljehani R, et al. Regulating AI-Based Medical Devices in Saudi Arabia: New Legal Paradigms in an Evolving Global Legal Order. Asian Bioethics Review. 21 June 2024. (Analysis of the SFDA's MDS-G010 guidance.) doi.org/10.1007/s41649-024-00285-6
  8. Australia Therapeutic Goods Administration. Consultation: Clarifying and strengthening the regulation of Artificial Intelligence (AI). Accessed July 2026. consultations.tga.gov.au/tga/clarifying-and-strengthening-the-regulation-of-ai/
  9. US FDA, Health Canada, and UK MHRA. Good Machine Learning Practice for Medical Device Development: Guiding Principles. October 2021. www.fda.gov/medical-devices/software-medical-device-samd/good-machine-learning-practice-medical-device-development-guiding-principles