Choose an assessment body by whether its legal entity and designated scope fit the product and route, then compare the expertise, process and lifecycle engagement your application needs. Scarlet, BSI and other bodies should not be ranked by reputation alone. The appropriate choice depends on the actual device, evidence and markets involved.
This article compares selection criteria and verified public eligibility information as checked on 13 October 2026. It does not report private quotations or a hands-on comparison of assessment quality. It is published by iatroX, whose clinical safety advisory services are distinct from the independent certification services being considered.
Establish the route before drawing up a shortlist
A founder should first specify the product's intended purpose, proposed market and likely assessment needs. The MHRA's intended-purpose guidance and market-access guidance provide relevant British starting points. Different routes and jurisdictions can require different assessment arrangements.
A management-system certification enquiry is not identical to a medical-device conformity enquiry. Nor is a UK approved-body entry interchangeable with an EU notified-body designation. Asking every organisation for a generic "AI certificate" is unlikely to produce comparable responses.
The shortlist should therefore begin with eligibility rather than a preferred brand. A well-known body whose scope does not fit the device is not the right partner for that assessment, however reassuring its name may appear in an investor presentation.
What the current UK list establishes
The MHRA list, updated on 17 July 2026 and checked on 13 October 2026, includes both of the following entities. The table summarises the listed categories; it does not replace checking the full designated scope for a particular device.
| UK legal entity | Identifier | Categories shown in the MHRA list | What still needs checking |
|---|---|---|---|
| BSI Assurance UK Ltd | 0086 | Medical devices, active implantable medical devices and in-vitro diagnostic medical devices under the relevant UK MDR parts | Detailed scope, proposed route and acceptance of the specific engagement |
| Scarlet NB UK Ltd | 8536 | Medical devices under Part II of the UK MDR | Detailed scope, proposed route and acceptance of the specific engagement |
These categories should not be used to infer relative clinical AI expertise or assessment quality. A broader category listing is not a score, and a narrower listing does not establish better specialisation for every software product.
For a European route, consult the European Commission's notified-body information and the relevant designation records. This UK table does not establish either company's EU scope.
Give every candidate the same product description
A consistent request helps prevent apparent differences that arise because the bodies received different information. Our proposed enquiry describes the intended purpose, users, population, functions, external dependencies, current development stage and target markets.
It should also state what evidence is complete, what remains planned and whether the product is expected to change during assessment. A start-up relying on an externally managed language model should describe that dependency rather than present a fixed-system diagram that omits it.
The good machine learning practice principles provide useful context for representative evidence, appropriate testing and human-AI performance. The enquiry should help the body understand the real assessment challenge, not conceal the difficult parts until after a quotation is accepted.
A reusable request-for-proposal outline
The following is an editorial template for comparable enquiries, not a form required by a regulator.
| Enquiry section | Information to supply | Response to request |
|---|---|---|
| Product scope | Intended purpose, users, population and functions | Eligibility and applicable assessment scope |
| Technical dependencies | Models, retrieval, connected tools and change exposure | Relevant assessment approach and information needs |
| Evidence readiness | Completed evaluations and known gaps | Expected stages and likely clarification needs |
| Engagement plan | Target milestones without assuming guaranteed approval | Capacity, communication and decision process |
| Lifecycle | Expected updates, surveillance and support needs | How subsequent work is assessed and charged |
| Commercial terms | Comparable assumptions across all candidates | Complete quotation, exclusions and foreseeable additional work |
The responses should be reviewed for comparability before the totals are compared. An initial quotation excluding later change assessment is not necessarily cheaper than a broader engagement once the same work is included.
Ask about expertise without demanding a predetermined answer
For a complex clinical AI product, ask how the body will obtain the relevant technical and clinical competence. The question is not whether it uses fashionable AI terminology, but whether it can interrogate the evidence and dependencies that matter to the device.
A useful response might explain how unfamiliar questions are escalated and how the assessment team is selected. It should not promise that the product will be accepted before the evidence is reviewed. The MHRA's account of approved-body designation emphasises competence and impartiality, not guaranteed commercial outcomes.
Scarlet publishes material on third-party LLM devices, which can inform an enquiry about that architecture. It is not a head-to-head result proving that Scarlet will assess every such device faster or better than another eligible body.
Compare the whole relationship, including changes
A manufacturer should understand how findings are communicated, how responses are handled and what happens when the planned product changes. It should also ask about surveillance, significant modifications and the circumstances requiring another assessment decision.
The joint predetermined change-control principles concern specified modifications and supporting evidence. A body offering an efficient process for such changes may be valuable, but the manufacturer must establish what is actually within scope and what remains subject to further review.
No comparable current private fee quotations or independently measured assessment-quality results are presented here. Consequently, this article does not name the cheapest, fastest or highest-quality provider. A selection based on unverified averages would be less useful than a disciplined comparison of the engagement the company actually needs.
A verdict by founder scenario
For a manufacturer with a defined route and submission-ready evidence, prioritise confirmed eligibility, appropriate capacity and clear communication. For a team with substantial clinical evidence gaps, advisory preparation may be the more immediate need before choosing a formal assessor.
For a product dependent on changing external models, give additional attention to the body's relevant experience and lifecycle arrangements, while applying the same questions to every candidate. For a multi-market business, verify each legal entity and route separately rather than assume one group-level relationship provides all necessary permissions.
As described on 13 October 2026, Insights by iatroX offers clinical safety and readiness support. That advisory work can help prepare a consistent product and evidence description; it does not replace a designated body's independent decision. The right partner is the one whose verified scope and complete engagement fit the product, not the one awarded an unsupported league-table position.
Frequently asked questions
Are Scarlet and BSI both listed as UK approved bodies?
Yes, the MHRA list checked on 13 October 2026 includes Scarlet NB UK Ltd and BSI Assurance UK Ltd. Their detailed scopes must still be checked against the proposed device and route.
Which certification body is cheapest or fastest for medical AI?
This article does not establish that ranking. Compare current quotations and clearly defined timelines for equivalent product scopes and complete engagements.
Should a founder choose the assessor before completing every study?
Early discussion can clarify eligibility and process, but it does not guarantee a favourable assessment. The manufacturer should be explicit about which evidence exists and which remains to be generated.