May 2026 was a high-density month for AI healthcare regulation, with significant activity across Canada, the UK, and the EU. Canada published new federal guidance addressing the governance of agentic AI systems, including considerations that may be relevant to federal health programs and other public-sector deployments. In the UK, the MHRA published draft pre-market regulations for medical devices and opened an industry-wide impact assessment consultation. The EU was the most active jurisdiction: a political agreement was reached on the AI Omnibus that extends the enforcement deadline for high-risk AI embedded in regulated products (including medical devices) to August 2028, while the Commission simultaneously published three sets of draft guidelines – on transparency obligations, high-risk AI classification, and a review of prohibited practices – signalling that compliance preparation remains urgent despite the extended timeline.
Disclaimer: This article is produced for educational and informational purposes only. It summarises publicly available regulatory updates and reflects our interpretation of emerging developments; it does not constitute legal, regulatory, financial, or professional advice, nor should it be interpreted as a statement of regulatory intent. Readers should consult qualified counsel before taking any action based on the information presented here.
Canada
Treasury Board of Canada Secretariat – Guide on the Use of Agentic Artificial Intelligence (May 22, 2026)
Category: Policy Guidance / Governance
The Treasury Board of Canada Secretariat (TBS) published its Guide on the Use of Agentic Artificial Intelligence on May 22, 2026, providing departments and agencies with practical direction for assessing, managing, and governing agentic AI systems. The guidance defines agentic AI as systems that can perceive and act on their environment autonomously, adapt their behaviour in response to changing inputs, and pursue goals over extended periods with limited human supervision. It also addresses multi-agent systems – composed of coordinated AI agents that break down tasks, collaborate, and use external tools – as well as auditability requirements and accountability obligations for federal institutions deploying these systems.
This guidance appears to be among the first federal Government of Canada publications focused specifically on the governance of agentic AI systems. The document provides information on oversight, accountability, and operational considerations associated with the use of autonomous, goal-directed AI within federal institutions. Its practical scope is broad: the guidance is directly applicable to AI tools being deployed in federal health programs, public health surveillance, and benefits administration. A notable feature of the guidance is its emphasis on auditability and record-keeping, including maintaining information about system actions, accessed resources, and operational changes. The guidance presents these practices as important governance and accountability mechanisms for federal institutions considering the use of agentic AI systems. Readers deploying or assessing agentic AI in federal healthcare settings should consult the original guidance and seek qualified regulatory and legal advice.
UK & EU
1) UK MHRA – Draft Medical Devices (Amendment) Regulations 2026 Published and Impact Assessment Consultation Opened (May 8-11, 2026)
Category: Medical Device Regulation / Legislation
On May 8, 2026, the MHRA published draft pre-market regulatory requirements – formally titled the Draft Medical Devices (Amendment) Regulations 2026 – on the World Trade Organisation notification portal, opening a comment period for WTO members. The draft regulations introduce new pre-market requirements for medical devices and in vitro diagnostic devices entering the Great Britain market, covering patient-centred requirements, patient safety, and access to innovative medical technologies. Simultaneously, the MHRA opened an impact assessment questionnaire for the MedTech industry, approved bodies, healthcare providers, and patients, running from May 11 to June 19, 2026.
These regulations directly govern how AI-enabled medical devices enter the GB market. The draft was developed following the MHRA’s 2024 consultation on medical device regulation routes to market, and represents the most substantive update to GB medical device pre-market rules in years. Those developing or distributing AI as a medical device in Great Britain should monitor this consultation closely and consider whether a submission to the impact assessment questionnaire is appropriate, in consultation with qualified regulatory advisors.
MHRA invites views on proposed changes to medical device regulation
2) EU – AI Omnibus Political Agreement Reached: High-Risk Medical Device AI Deadline Extended to August 2028 (May 7, 2026)
Category: Legislation / Timeline Change – Direct Healthcare AI Impact
On May 7, 2026, the European Parliament and the Council of the EU reached a political agreement on the AI Omnibus – the package of targeted simplification amendments to the EU AI Act. The most consequential outcome for healthcare AI is a formal extension of the enforcement deadline for high-risk AI systems embedded in regulated products, including medical devices governed by EU MDR and IVDR, to August 2, 2028. For other high-risk AI systems (covering biometrics, critical infrastructure, employment, and education), the deadline is December 2, 2027. The agreement also formally bans so-called “nudification” AI applications.
For healthcare AI developers, the 2028 enforcement date for product-embedded AI provides meaningful operational runway. However, the Commission’s simultaneous issuance of draft guidelines on transparency, classification, and prohibited practices (see below) is a clear signal that compliance preparation is expected to proceed regardless of the extended deadline. Readers should consult qualified legal counsel on how the revised timeline interacts with their specific MDR/IVDR obligations and existing regulatory programmes.
Artificial intelligence: Council and Parliament agree to simplify and streamline rules
3) EU Commission – Draft Guidelines on Transparency Obligations Under Article 50 of the AI Act (May 8, 2026)
Category: Policy and Legislation / Compliance Guidance
The European Commission’s AI Office published draft guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act, opened for stakeholder consultation on May 8, 2026. Article 50 covers four distinct transparency obligations: (1) providers of AI systems that interact directly with natural persons must inform those persons they are interacting with an AI; (2) providers of systems generating or manipulating synthetic audio, image, video, or text must implement machine-readable marking and detection solutions; (3) deployers of emotion recognition or biometric categorisation systems must inform exposed individuals of the system’s operation; and (4) deployers generating deep fakes or AI-manipulated text published to inform the public on matters of public interest must disclose the artificial origin of that content. These obligations apply from August 2, 2026.
The guidelines are particularly relevant to healthcare AI in several respects. The draft guidelines indicate that AI systems interacting directly with natural persons may fall within Article 50(1). Depending on their design and use case, this could include certain patient-facing chatbots, virtual assistants, or similar healthcare applications. Generative AI systems producing synthetic medical images – noted explicitly in the guidelines as falling within Article 50(2) – must implement technically feasible marking and detection solutions that are effective, interoperable, robust, and reliable. Emotion recognition systems, which are also classified as high-risk under the AI Act, carry concurrent transparency obligations under Article 50(3). The guidelines note that a single AI system may trigger multiple obligations simultaneously, potentially engaging the responsibilities of both providers and deployers.
On enforcement, providers and deployers that do not comply with Article 50 may be fined up to €15,000,000 or, if the offender is an undertaking, up to 3% of its total worldwide annual turnover for the preceding financial year, whichever is higher. Adherence to a code of practice deemed adequate by the AI Office may be used to demonstrate compliance and may be treated as a mitigating factor in any penalty assessment. These guidelines are non-binding and remain in draft for consultation; authoritative interpretation rests ultimately with the Court of Justice of the EU. Affected providers and deployers should consult the full guidelines and seek qualified legal advice on their specific obligations.
4) EU Commission – Draft Guidelines on the Classification of High-Risk AI Systems (May 19, 2026)
Category: Policy and Legislation / Classification Guidance
The Commission published draft guidelines on May 19, 2026 on the classification of high-risk AI systems under the AI Act. These guidelines are intended to assist providers and deployers in determining whether their AI systems meet the threshold for high-risk classification under Article 6, which triggers a substantially more demanding set of obligations including conformity assessments, technical documentation, human oversight requirements, and registration. The guidelines are in draft for stakeholder consultation and remain non-binding pending finalisation.
For healthcare AI specifically, correct classification is a critical upstream determination: AI systems integrated into medical devices regulated under MDR or IVDR are subject to the high-risk pathway, and misclassification carries both compliance and potential regulatory risk. Providers and deployers of AI systems in healthcare settings may wish to review the draft guidelines carefully in conjunction with qualified legal and regulatory counsel.
Draft Commission guidelines on the classification of high-risk AI systems
5) EU Commission – Report on the Review of Prohibitions and High-Risk AI (May 22, 2026)
Category: Policy and Legislation / Prohibited Practices Review
The Commission published a report on May 22, 2026 reviewing the scope of prohibited AI practices under Article 5 of the AI Act and examining the boundary between prohibited and high-risk classifications. Article 5 prohibitions represent the absolute ceiling of the AI Act’s risk-based framework – AI applications whose risks are considered unacceptable regardless of mitigating measures. These include, among others, the use of emotion recognition systems in workplace and educational settings, and certain biometric categorisation practices.
For healthcare AI developers, this report is relevant both as a direct compliance reference and as an indicator of regulatory direction. AI systems operating in clinical environments that involve biometric data, emotional state inference, or behavioural monitoring may need to be assessed against the prohibition boundary – not only against the high-risk threshold. Readers should consult qualified legal counsel on how the prohibition categories interact with their specific products and use cases.
Cross-Cutting Themes
Disclaimer: The cross-cutting themes below are analytical observations based on the sources cited in this article and reflect our interpretation of emerging regulatory and policy developments. They do not constitute legal, regulatory, or professional advice, nor should they be interpreted as statements of regulatory intent. Readers may wish to consult qualified advisors regarding the application of any regulatory requirements to their specific circumstances.
1) Agentic AI is entering formal governance frameworks.
Canada’s Guide on Agentic AI is the first national-level framework in the country to address autonomous, goal-directed AI systems directly. The guidance’s emphasis on auditability and accountability reflects a broader policy focus on documenting and overseeing increasingly autonomous AI systems. This is likely to be a continuing area of development across jurisdictions, and those deploying agentic AI in health-adjacent federal programs in Canada should treat this guidance as establishing the current governance floor.
2) The EU’s extended high-risk deadline does not reduce compliance urgency.
The AI Omnibus agreement’s extension of the enforcement deadline for high-risk AI in regulated products to August 2028 provides operational runway, but the Commission’s concurrent publication of multiple draft guidelines may indicate an intention to provide additional interpretive guidance while implementation timelines continue to evolve. The guidelines on transparency, classification, and prohibited practices are all open for stakeholder input and will be finalised in the lead-up to the 2028 deadline. For healthcare AI developers, the window between now and 2028 may be best understood as a structured preparation period, not a deferral.
3) Transparency obligations are now being operationalised in detail.
The Article 50 draft guidelines represent the most granular treatment yet of what AI transparency means in practice – covering disclosure format, timing, content, technical marking standards, and enforcement pathways. For patient-facing AI tools in healthcare (chatbots, virtual assistants, diagnostic support systems with interactive outputs), the guidelines set out specific expectations that go beyond a general disclosure principle. Providers and deployers may wish to assess their current disclosure practices against the draft guidance and begin gap analysis work, accepting that the final guidelines may differ from the current draft.
4) The GB and EU regulatory tracks for AI as a medical device are diverging.
The MHRA’s draft Medical Devices (Amendment) Regulations 2026 represent an independent GB pre-market framework developing in parallel to EU MDR/IVDR. With the EU AI Act’s high-risk deadline for medical device AI now set at August 2028, and the GB framework still in draft consultation, Organisations seeking market access in both jurisdictions may need to monitor developments under both regulatory frameworks as they continue to evolve independently. Regulatory planning for AI-enabled medical devices seeking both GB and EU market access may benefit from early dual-track assessment in consultation with qualified advisors.
5) Classification decisions are becoming a critical upstream risk.
The Commission’s draft guidelines on high-risk AI classification underscore a point that is easy to underestimate: the determination of whether an AI system is high-risk is not self-evident, and an incorrect classification carries significant downstream compliance and enforcement exposure. In healthcare, where AI systems may interact with regulated product frameworks (MDR/IVDR), biometric data, and clinical decision-making, the classification boundary warrants careful assessment. The draft guidelines, once finalised, will provide the primary interpretive reference – but in the interim, affected organisations should not defer this analysis.
Key Considerations for Regulatory Alignment
Disclaimer: This checklist is provided for general informational purposes only and does not constitute legal, regulatory, or professional advice; organizations should consult with their legal and compliance departments to ensure adherence to specific jurisdictional requirements.
For Founders & Business Owners
1) Assess whether your AI products fall within the Article 50 transparency scope.
The Article 50 obligations apply from August 2, 2026 and cover interactive AI systems, generative AI systems, emotion recognition, biometric categorisation, and deep fake or public-interest AI text outputs. For patient-facing products – chatbots, voice assistants, diagnostic interfaces – the question of whether and how to disclose the AI nature of interactions is now an obligation described in the draft guidance, not a design choice. Early legal review of your disclosure practices against the draft guidelines may be relevant to consider.
2) Factor the 2028 high-risk deadline into your regulatory roadmap – but don’t treat it as a pause.
The AI Omnibus extension gives developers of AI embedded in medical devices until August 2028 to comply with the high-risk AI rules. However, preparatory obligations – including technical documentation, conformity assessment planning, and human oversight design – are complex and time-consuming. Organisations that begin implementation planning now are likely to be better positioned than those who treat the 2028 date as the start, rather than the end, of that work.
3) If operating agentic AI in Canadian federal health contexts, review the TBS guidance promptly.
Canada’s Guide on Agentic AI establishes auditability and accountability requirements that apply to autonomous AI operating in federally adjacent healthcare settings. If your organisation deploys or is considering deploying agentic AI tools in federal health programs, public health analytics, or benefits administration, the guide sets out what governance structures are now expected. Consulting qualified regulatory and legal advisors on implementation is advisable.
4) Consider engaging with the MHRA’s impact assessment questionnaire if operating in the GB market.
The MHRA’s impact assessment consultation for the Draft Medical Devices (Amendment) Regulations 2026 runs until June 19, 2026. For AI medical device developers with GB market ambitions or existing GB approvals, this consultation represents an opportunity to shape the pre-market framework before it is finalised. Organisations may wish to review the draft regulations with qualified regulatory counsel and consider whether a formal response is appropriate.
5) Be aware that the GB and EU regulatory tracks for medical devices are developing independently.
Organisations pursuing dual GB and EU market access for AI-enabled medical devices now face two distinct and potentially divergent pre-market frameworks. Early dual-track regulatory planning – with qualified advisors familiar with both regimes – is likely to be more efficient than sequential engagement.
For Compliance & Regulatory Specialists
1) Begin or advance gap analysis against the Article 50 draft transparency guidelines.
The draft guidelines provide the most operationally detailed treatment to date of the Article 50 obligations. For in-scope systems, a structured gap analysis comparing current disclosure practices against the guidelines’ requirements – covering format, timing, content, and technical marking standards – will be necessary regardless of whether the final guidelines differ from the draft. Starting this work now allows for iterative adjustment as the guidelines are finalised ahead of the August 2, 2026 application date.
2) Review AI system classification decisions in light of the draft high-risk classification guidelines.
The draft guidelines on high-risk AI classification published May 19 will provide the primary reference for determining whether a given AI system triggers the full high-risk compliance pathway. For healthcare AI, where the boundary between high-risk and non-high-risk is commercially and operationally significant, classification assessments should be documented, reviewed by qualified legal counsel, and updated as the guidelines are finalised. This is particularly relevant for AI systems that sit at the interface of the AI Act and MDR/IVDR.
3) Assess marking and detection technical readiness for Article 50(2) compliance.
The Article 50(2) obligation – requiring machine-readable marking and corresponding detection solutions for synthetic audio, image, video, and text outputs – applies from August 2, 2026. The draft guidelines clarify that no single technique currently meets all four requirements (effectiveness, reliability, robustness, interoperability) simultaneously, meaning a combination of techniques will generally be required. Technical teams should be engaged now to assess what solutions are feasible given current system architecture, and compliance counsel should be involved in determining whether the resulting solution meets the legal standard.
4) Map the AI Omnibus timeline extension against existing MDR/IVDR compliance programmes.
The extension of the high-risk AI rules deadline to August 2028 for product-embedded AI (including medical devices) may create an opportunity to align AI Act compliance work with existing MDR/IVDR regulatory programmes rather than running them as entirely parallel tracks. Those managing compliance programmes across both frameworks may wish to assess whether integration points – technical documentation, conformity assessment, post-market surveillance – can be structured to reduce duplication. Qualified legal counsel should be involved in any such alignment assessment.
5) Review the agentic AI governance requirements for any Canadian federal health program deployments.
The TBS Guide on Agentic AI introduces specific requirements around auditability, multi-agent governance, and accountability that may exceed what existing AI governance frameworks cover. Compliance professionals supporting organisations with deployments in Canadian federal health programs should cross-reference current governance documentation against the guide – particularly the requirements for time-stamped audit logs, human oversight thresholds, and accountability chain documentation for multi-agent systems. Qualified regulatory counsel should be engaged for any detailed compliance assessment.
Sources
- Guide on the Use of Agentic Artificial Intelligence (May 22, 2026)
https://www.canada.ca/en/government/system/digital-government/digital-government-innovations/responsible-use-ai/guide-use-agentic-artificial-antelligence.html - MHRA invites views on proposed changes to medical device regulation (May 8, 2026)
https://www.gov.uk/government/news/mhra-invites-views-on-proposed-changes-to-medical-device-regulation - Artificial intelligence: Council and Parliament agree to simplify and streamline rules (May 7, 2026)
https://www.consilium.europa.eu/en/press/press-releases/2026/05/07/artificial-intelligence-council-and-parliament-agree-to-simplify-and-streamline-rules/ - Draft guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act (May 8, 2026)
https://digital-strategy.ec.europa.eu/en/library/draft-guidelines-implementation-transparency-obligations-certain-ai-systems-under-article-50-ai-act - Draft Commission guidelines on the classification of high-risk AI systems (May 19, 2026)
https://digital-strategy.ec.europa.eu/en/library/draft-commission-guidelines-classification-high-risk-ai-systems - Report on the review of prohibitions and high-risk AI (May 22, 2026)
https://digital-strategy.ec.europa.eu/en/library/report-review-prohibitions-and-high-risk-ai
