OpenAI's 1 September announcement offers a genuinely useful starting point for a different exercise: designing what a UK equivalent would actually need to contain, not as a prediction of what OpenAI or any specific company will build, since nothing of this kind has been announced, but as a proposal worth stating explicitly, because articulating the UK stack in full makes clear exactly how much more than source substitution the task would involve.
The proposed UK connector stack
Clinical guidance: NICE guidance and NICE CKS as the core national reference, SIGN for Scotland-specific guidance, national clinical pathways more broadly, and specialist-society guidance where appropriately licensed for this kind of use, the layer that answers where a treatment sits in a UK-specific pathway, a question no US source in OpenAI's current stack addresses at all.
Medicines: UK SmPCs and PILs through emc as the product-information layer, MHRA Drug Safety Updates and safety alerts or recalls as the live regulatory-communication layer, dm+d medicine identifiers as the UK's own terminology-standardisation layer, and Specialist Pharmacy Service advice for the practical, complex-case questions that fall between guideline and product information, the medicines stack this cluster's dedicated DailyMed and RxNorm comparisons map against their nearest US equivalents throughout.
Trials and research: PubMed, genuinely international already and requiring no UK-specific substitute, ClinicalTrials.gov likewise, alongside ISRCTN as the UK and international trial registry with particular UK relevance, NIHR research infrastructure specifically, and UK clinical-trial registries and recruitment information more broadly, extending the research layer beyond the purely bibliographic into UK-specific trial access.
Organisations and professionals: the GMC, NMC, GPhC and HCPC registers as the UK's professional-identification layer, the equivalent function the NPI Registry serves in the US stack, alongside NHS ODS information for organisational identification, CQC information for the UK's quality and inspection layer, and NHS service directories for locating actual care provision.
Local operational context: formularies, antimicrobial policies, referral pathways, shared-care agreements and local diagnostic thresholds, the layer that makes any national guidance operationally usable at the level of an individual trust or practice, and the layer with no equivalent anywhere in OpenAI's current nine-source stack, since it is inherently local rather than national in scope.
The core argument
A UK source plugin built this way should not merely translate American content into ostensibly equivalent UK categories. It should represent the regulatory, medicines, commissioning and clinical-pathway architecture UK practice actually runs on, which is a structurally different task from source substitution, the same finding this cluster's dedicated US-first gap analysis reaches from the opposite direction: the mapping between US and UK sources is not one-for-one because the underlying institutions are not one-for-one, and a genuine UK connector stack has to be built around UK institutions from the ground up rather than assembled by finding the nearest American-source equivalent for each category.
Why this speculative exercise is worth doing explicitly
Naming the full stack in this much detail does two things at once: it demonstrates concretely how much more the task involves than headline-level comparison suggests, five categories, at least twenty distinct source types, spanning national guidance, medicines regulation, research infrastructure, professional registration and, critically, local operational context no national source can ever fully replace. And it functions as a transparent roadmap for exactly the kind of source architecture iatroX is already building toward, stated plainly as aspiration and direction rather than as a claim about capability this platform does not yet fully possess across every category listed.
The speculative element, stated clearly
OpenAI has not announced NICE, emc or any other UK-specific connector, and nothing in this article should be read as reporting an actual OpenAI product plan. This is iatroX's own proposal for what the UK equivalent of the 1 September announcement would need to contain, offered because the exercise clarifies both the scale of the UK gap and the shape of the work worth doing to close it.
Frequently asked questions
Could OpenAI plausibly build this stack itself?
Technically nothing prevents it, and doing so would require the kind of deep, UK-institution-specific partnership and licensing work this article's stack implies, a genuinely large undertaking distinct from simply adding more API connections to the existing US-oriented plugin.
Which category in this proposed stack is hardest to build well?
Local operational context, plausibly, since formularies, antimicrobial policies and shared-care agreements vary genuinely by individual trust and practice, meaning this layer cannot be solved with a single national data source the way clinical guidance or professional registration largely can.
Does iatroX currently cover every category in this proposed stack?
Not uniformly across every category listed, and the platform's current strength concentrates specifically in clinical guidance and medicines information, with the fuller local-operational-context layer representing genuine ongoing direction rather than an already-complete capability.