Weekly News and Updates (Nov 8 – 21, 2025)

by | Nov 23, 2025 | AI News & Updates | 0 comments

Between 8-21 November 2025 regulators and international bodies emphasised moving from principles to practice: the EU launched COMPASS-AI to operationalise safe clinical AI; the UK (MHRA) published AI Airlock pilot outputs and announced AI drug-safety projects; the FDA continued to press lifecycle and real-world evaluation (publishing meeting materials and continuing workshops on modelling & AI in drug development); WHO regional activity reinforced governance & capacity needs; and national cyber bodies flagged AI-specific threats to health infrastructure.

 

 

 

 

 

 

 

 

 

 

Canada

 

 

1) Federal AI strategy activity – AI Strategy Task Force and national “sprint” (Oct → Nov 2025; Task Force outputs shared in November)

 

The Government of Canada (ISED) launched an AI Strategy Task Force and a national AI “sprint” (Oct 1-31, 2025). The Task Force convened public engagement through October and indicated it would share the bold, practical ideas gathered in early November 2025 as it shapes Canada’s next AI strategy. While not health-only, the federal effort explicitly flags health-sector AI as a priority area for coordination, funding and pilot activity.

ISED – Help define the next chapter of Canada’s AI leadership (AI Strategy Task Force & sprint)

 

How it applies to AI in healthcare: expect increased federal coordination and potential funding or sandbox opportunities for clinical pilots – align proposals and governance documentation (safety, equity, data stewardship) to the national consultation themes.

 

 

 

 

2) Health Canada planning & related federal guidance (operational context through Nov 2025)

 

Health Canada’s 2025-26 departmental plan and federal guidance on responsible AI use in government remain the operative policy context for federal health regulation and procurement decisions during the Nov 8-21 window. These documents emphasise safety, interoperability and risk-based oversight of emerging tech in health services.

Health Canada – Departmental Plan 2025–26
Government of Canada – Artificial Intelligence (ecosystem & guidance pages)

 

How it applies to AI in healthcare: vendors and health organisations bidding for federal pilots or procurement should prepare departmental-aligned evidence packages (safety, equity, interoperability) and be ready to demonstrate alignment with federal responsible-AI guidance and procurement expectations.

 

 

 

 

 

 

 

 

United States

 

 

1) FDA – Digital Health Advisory Committee materials & generative AI deliberations (early November 2025)

 

The FDA’s Digital Health Advisory Committee met to discuss generative-AI-enabled digital mental health devices (meeting materials, discussion questions, and a 24-hour meeting summary are publicly posted). The Committee focused on benefits, risks, clinical evidence expectations, and lifecycle/post-market monitoring for generative AI devices. These materials were published by FDA in early November and continued to shape regulatory expectations through the Nov 8-21 window.

FDA – DHAC meeting announcement (Nov 6, 2025)
FDA – DHAC meeting agenda / materials (Nov 6, 2025)
FDA – 24-hour summary of DHAC (Nov 6, 2025)

 

How it applies to AI in healthcare: regulators expect robust premarket evidence and explicit post-market monitoring plans (drift, hallucination tracking, user-interaction logs) for generative AI medical devices – especially for mental-health patient-facing tools. Build lifecycle evidence and monitoring dashboards into your regulatory strategy.

 

 

 

 

2) FDA CRCG workshop – Modelling & AI in generic drug development (Oct 15-16 outputs remain relevant)

 

The FDA Center for Research on Complex Generics (CRCG) held a public workshop on modelling and AI in generic drug development; the CRCG materials and agenda are posted and are continuing points of regulatory engagement into November (e.g., emphasis on model validation, provenance and lifecycle controls for modelling in regulatory submissions).

FDA – CRCG Workshop: Modeling & AI in Generic Drug Development (Oct 15-16, 2025)

 

How it applies to AI in healthcare: if using AI for formulation, bioequivalence modelling, or lifecycle analytics, be prepared to submit model provenance, reproducible validation plans, and continuous monitoring measures as part of regulatory dossiers.

 

 

 

 

 

 

 

 

European Union & United Kingdom

 

 

1) European Commission – Launch of COMPASS-AI (21 Oct 2025; operational outputs active through Nov)

 

The European Commission launched COMPASS-AI (a flagship initiative under the EU’s Apply AI strategy) to operationalise AI in healthcare via multidisciplinary expert communities, pilot clinical deployment guidelines, and a knowledge-sharing digital platform. The official press and policy pages were published by the Commission and remained reference documents through November as COMPASS-AI moved to pilot and stakeholder engagement phases.

European Commission – Commission launches flagship initiative to increase use of AI in healthcare (21 Oct 2025)
EU – Artificial Intelligence in Health (policy hub)

 

How it applies to AI in healthcare: COMPASS-AI emphasises pilot validation, common evaluation metrics and clinician literacy. Align clinical validation, interoperability and ethics-by-design workstreams to COMPASS-AI outputs to improve procurement and cross-border acceptance in the EU.

 

 

 

 

2) UK – MHRA AI Airlock sandbox programme report & new AI drug-safety projects (published Oct → mid-Nov activity)

 

The MHRA published the AI Airlock Sandbox pilot programme report on 16 October (report PDF), and also announced a government-backed programme using AI and anonymised NHS data to predict drug-interaction side effects (announced 22 Oct). The Airlock report contains practical recommendations on explainability, synthetic data, fidelity checks and post-market surveillance – all being referenced across the sector during Nov 8-21.

MHRA – AI Airlock Sandbox Pilot Programme Report (16 Oct 2025; PDF)
MHRA – Side effects from drug interactions to be predicted by AI (22 Oct 2025)

 

How it applies to AI in healthcare: the Airlock outputs function as de-facto evaluation checklists for explainability, hallucination testing and real-world performance – adopt these test suites and transparency elements when planning UK pilots or procurement.

 

 

 

 

 

 

 

 

Rest of the world

 

 

1) WHO – Regional committee outputs (Western Pacific Session 76; Oct 20-24 materials but referenced through Nov)

 

WHO Western Pacific Regional Committee session documents (session 76) emphasised AI governance, capacity building, data stewardship and equity in deployment. WHO’s ongoing “Harnessing AI for Health” program pages continue to provide technical guidance and capacity-building priorities used by Member States into November.

WHO – Seventy-sixth session of the Western Pacific Regional Committee (20-24 Oct 2025)
WHO – Harnessing artificial intelligence for health (program page)

 

How it applies to AI in healthcare: for international deployments and low/middle-income settings, align pilots with WHO governance and capacity building recommendations; this supports donor funding and cross-jurisdictional acceptance.

 

 

 

 

2) Australia – Annual Cyber Threat Report (2024-25) highlighting AI-enabled threats (published Oct 2025; implications through Nov)

 

The Australian Cyber Security Centre (ACSC) / ASD published the Annual Cyber Threat Report 2024-25 (released mid-October), explicitly flagging AI as both a defensive enabler and an emerging tool for attackers targeting critical infrastructure – including healthcare. The report recommends treating AI models and training data as high-value assets and planning adversarial-resilience measures.

ASD/ACSC – Annual Cyber Threat Report 2024-25 (published Oct 2025)

 

How it applies to AI in healthcare: every clinical AI deployment must have a tailored threat model (data poisoning, model extraction, hallucination abuse), hardened access controls, and incident response integrated with clinical-safety escalation paths.

 

 

 

 

 

 

 

 

Cross-Cutting Themes Across Jurisdictions (observed during 8-21 Nov 2025)

 

  1. From principles to pilots: EU COMPASS-AI and MHRA Airlock emphasise operational pilots, shared evaluation metrics and knowledge platforms so regulators and health systems can converge on practical validation approaches.
  2. Total-lifecycle & continuous monitoring: FDA DHAC materials and CRCG workshop outputs underscore lifecycle evidence, continuous monitoring, drift detection and retraining governance as central regulatory expectations.
  3. Regulator-sponsored safety tools: MHRA’s drug-interaction AI projects show regulators are directly funding tools to improve pharmacovigilance – implying higher data-quality and validation expectations for vendor solutions used in safety surveillance.
  4. Security & data governance: national cyber reports (e.g., ACSC/ASD) plus WHO guidance stress AI security, data stewardship, equity and capacity building as integral to safe cross-border deployment.

 

 

 

 

 

 

 

 

Immediate, concrete checklist for health organisations & vendors (actionable during Nov 8-21, 2025)

 

  • Map applicable regimes: For each product/feature, map EU (COMPASS-AI outputs + AIA compliance), UK (MHRA sandbox outputs), FDA (GMLP & lifecycle expectations), Canada (AIDA context & federal guidance), local privacy laws and WHO guidance.
  • Prepare an evidence dossier: model card, training data summary, demographic/subgroup performance, bench & real-world validation reports, explainability statement, hallucination tests (adopt MHRA Airlock test concepts).
  • Implement lifecycle controls: versioning, drift detection, retraining governance, continuous-monitoring dashboards, and clinical safety KPIs tied to incident escalation.
  • Security & threat modelling: treat models & training data as critical assets – access controls, encryption, logging, supply-chain due diligence, and adversarial-resilience testing per ACSC guidance.
  • Update privacy & PIA: update Privacy Impact Assessments for inference/telemetry flows, RWD linkage and de-identification measures.
  • Procurement & contracting: require vendor transparency (model provenance, monitoring SLAs, retraining triggers, audit logs, regulatory attachments) and pilot evidence for clinical use.
  • Engage regulators & sandboxes early: use MHRA Airlock, COMPASS-AI pilots, FDA engagement forums and other national sandboxes to validate evidence and receive formative regulatory feedback.

 

 

 

 

 

 

 

 

Sources (original, primary documents and official pages)

 

Written by Grigorii Kochetov

Cybersecurity Researcher at AI Healthcare Compliance

Read more

Prohibited AI Systems Under the EU AI Act

Prohibited AI Systems Under the EU AI Act

The European Union’s Artificial Intelligence Act (EU AI Act) establishes the world’s first comprehensive legal framework for governing artificial intelligence. It divides AI systems into four categories based on their potential impact on safety and fundamental rights...

read more
Practical impacts of using AI in Healthcare

Practical impacts of using AI in Healthcare

Artificial Intelligence (AI) is transforming healthcare systems globally - enhancing diagnostics, improving patient outcomes, optimizing workflows, and reducing costs. However, its adoption also brings challenges around data integrity, equity, and ethical use. Below...

read more
Weekly News and Updates (Sept 19–25, 2025)

Weekly News and Updates (Sept 19–25, 2025)

This post will begin our new weekly updates that will cover the most recent developments in AI governance and regulations, with a particular focus on how these changes affect AI in healthcare. We begin with updates from Canada — including privacy enforcement actions,...

read more
Healthcare AI Impact:Speed and Efficiency

Healthcare AI Impact:Speed and Efficiency

AI integrations are accelerating healthcare like never before. From cutting radiology wait times to reducing the hours physicians spend on documentation, AI is proving to be one of the most powerful efficiency drivers in medicine. This article explores how much time...

read more