Guidance, standards and papers

FDA denies a partial 510(k) exemption for radiology CADe, CADx and CADt software. These devices still need 510(k) clearance.

UK commission recommends a proportionate, lifecycle-based framework for regulating AI in health care, including staged authorisation.

EU ethics group argues neuro-AI should be governed as shared infrastructure, with safeguards for mental privacy and integrity.

FDA discussion paper proposing a risk framework and premarket approach for generative AI devices, with a request for feedback.

CMS notice proposing a RAPID coverage pathway that aligns Medicare evidence expectations with FDA review for new devices.

Meta-analysis of 42 studies on FDA-cleared and CE-marked AI for breast screening, pooling accuracy by clinical role.

FDA's updated final human factors guidance, replacing the 2016 version and aligned with the QMSR and ISO 14971:2019.

Draft FDA and industry commitments for MDUFA VI: review goals for Pre-Subs, 510(k)s, De Novos and PMAs for FY 2028 to 2032.

MDIC best practices, co-authored with FDA staff, for planning, running and reporting device penetration tests.

MHRA draft amendments setting new UK pre-market requirements, notified to the WTO in May 2026.

NESTcc draft tool for rating real-world data sources on quality, completeness and fitness for regulatory use.

FDA guidance on which human factors information to include in a marketing submission, by risk category.

IMDRF draft framework defining an eight-step life cycle for AI-enabled medical devices, from planning to decommissioning.

MITRE considerations for normalizing SBOM data so component and vulnerability records can be compared and trusted.

MITRE paper on adapting device cybersecurity risk analysis to AI, cloud and connected-care technologies.

CMS details for the ACCESS Model, including payment amounts and performance targets for technology-enabled care.

The program manual FDA investigators use to inspect device manufacturers, updated for the QMSR.

FDA guidance on the cybersecurity design, documentation and SBOM content expected in premarket submissions.

CDRH's review of 2025: device authorizations, digital health and AI activity, and program milestones.

FDA guide to using clinical data from outside the US in device submissions.

FDA request for information on its Foundational Innovation and RAPID Engagement (FIRE) program.

MDCG guidance setting out how breakthrough devices can get early dialogue and prioritised review in the EU.

FDA final guidance on when real-world data is relevant and reliable enough to support device regulatory decisions. Replaces the 2017 version.

FDA pilot on technology-enabled meaningful patient outcomes for digital health devices.

Materials for FDA's advisory committee meeting on generative AI-enabled digital mental health devices.

Real-world study measuring how AI triage changes radiologist report turnaround times.

NEJM AI case study of how APPRAISE-HRI, an AI tool that triages trauma casualties by hemorrhage risk, was developed and cleared by FDA.

Paper on how hallucination in AI systems applies to medical devices and how to evaluate and mitigate it.

USPTO open letter and memo on institution authority for patent trial proceedings.

FDA final guidance on a risk-based, least-burdensome approach to validating software used in production and the QMS.

European Commission guide to the EU Data Act and its data-sharing obligations for connected products, including devices.

Cosm's checklist of what a software as a medical device 510(k) needs, section by section.

MedTech Europe guide to clinical evidence and performance evaluation requirements under the IVDR.

NIST standard for lightweight cryptography suited to low-power devices, based on the Ascon family.

CISA's updated draft of the minimum data fields every SBOM should include.

FDA De Novo classification order for ArteraAI's AI-based prostate cancer prognostic software.

First MDCG guidance on how the EU AI Act applies to AI-enabled medical devices and IVDs, including high-risk classification.

EU report on generative AI: technology trends, economic impact and policy considerations.

MDCG guidance on obligations when medical device software apps are distributed through online platforms and app stores.

JAMA viewpoint by FDA leadership setting out the agency's new priorities.

Study using physics-informed deep learning to estimate 3D aortic blood flow from anatomy alone.

FDA De Novo order for Clairity Breast, AI that predicts five-year breast cancer risk from a mammogram.

FDA's quarterly report on review times and goals under MDUFA V.

FDA draft guidance on documentation and lifecycle practices for AI-enabled devices in marketing submissions.

EU study mapping near-eye XR applications and services, with market trends and policy options for health and other sectors.

FDA guidance on showing that computational models used as evidence in a submission are credible.

FDA guidance on biocompatibility evaluation under ISO 10993-1 within a risk management process.

FDA guidance on the software documentation to include in a premarket submission, by documentation level.

FDA draft guidance, now superseded, on human factors content for marketing submissions.

FDA guidance on designing clinical performance studies for radiology CADe devices.

FDA guidance on what to include in a 510(k) for radiology computer-assisted detection devices.

FDA guidance on software functions that transfer, store or display device data and where FDA does not enforce requirements.

FDA guidance on the technical performance data needed for quantitative imaging functions in radiological devices.

IMDRF's shared vocabulary for machine learning-enabled medical devices.

MDIC and MITRE playbook teaching practical threat modeling for medical device teams.

FDA guidance on the Special 510(k) route for changes to your own cleared device, and when design controls support it.

FDA guidance on declaring conformity to recognized standards in premarket submissions.

FDA guidance on monitoring, assessing and fixing cybersecurity vulnerabilities in marketed devices.

FDA's original 2016 human factors guidance, now replaced by the August 2026 version.

NIST's method for information security risk assessment, widely used for device threat and risk analysis.

A one-page overview of Cosm's regulatory, quality and product development services.