For an NHS Trust, an AI scribe certified at Class IIa means an independent Approved Body has assessed the product's safety and quality management for a higher-risk intended purpose, rather than the manufacturer simply self-declaring it. That shifts part of the clinical-safety assurance from your local committee to an external assessment, which can reduce some diligence burden. It does not remove your local responsibilities, and certification is never a guarantee of zero error. Here is what it actually means for procurement.
Key takeaways
- Class IIa requires independent certification of the manufacturer's quality system and technical documentation.
- That external assessment lets your local clinical-safety work lean on it, but does not replace it.
- Device conformity and product liability sit with the manufacturer; deployment and use responsibilities sit with the Trust and clinician.
- The efficiency gain is "assessed once, relied on many times" across Trusts, rather than each reassessing from scratch.
- Certification is not zero-error assurance, so post-market surveillance and human verification remain essential.
Who is allowed to certify a device, and on what evidence?
The class determines who does the checking. A Class I device is self-declared: the manufacturer assesses its own conformity and registers with the MHRA, with no external body reviewing the file. A Class IIa device requires an independent Approved Body (under the UK's UKCA regime) or notified body (under EU MDR, for CE marking) to carry out a conformity assessment. That assessment audits the manufacturer's quality management system and examines technical documentation and clinical evaluation before a certificate is issued. So at Class IIa, an external, accredited body has looked at the evidence, which is the substantive difference from self-declaration.
How does external certification shift clinical-safety judgement?
In the NHS, deploying clinical software involves local clinical-safety work under standards such as DCB0160, alongside the manufacturer's own DCB0129 responsibilities. When a product is Class IIa certified, part of the safety and quality assurance has been independently validated, which your local clinical-safety case can reference rather than reconstruct. What it does not do is discharge your local duties. You still need a named Clinical Safety Officer, a local hazard log and safety case, and a Data Protection Impact Assessment. The certification raises the floor of assurance under the product; it does not lift the responsibility off the Trust.
Where does product liability sit?
Broadly, responsibility follows the boundary between the product and its use. The manufacturer is responsible for the device conforming to its certified intended purpose and for product-level liability. The Trust is responsible for appropriate deployment, governance, consent processes and training. The clinician is responsible for reviewing and approving outputs before they enter the record. Certification does not transfer use-side responsibility to the manufacturer, so using a certified tool outside its intended purpose, or without human verification, puts the risk back on the deployer and the clinician.
What does "diligence performed once, used many times" mean?
This is the practical value of the regime. When an Approved Body certifies a device, that assessment is done once and can be relied upon by every Trust considering the product, rather than each Trust independently trying to assess the manufacturer's quality system from the outside. It does not remove local DTAC, information governance or clinical-safety steps, but it means the deepest technical assessment has already been done by an accredited body. For a procurement team, that is a meaningful reduction in duplicated effort and a more consistent baseline across the system.
What are the limits?
Three limits are worth stating plainly. First, certification is not a guarantee of zero error; it certifies that a process and intended purpose meet a standard, not that outputs are always correct. Second, a certificate is scope-bound: it covers a specific product and intended purpose, so check that your use matches. Third, if the vendor changes features, the intended use may shift and trigger reclassification, so certification is a point-in-time status that needs monitoring. This is why post-market surveillance, incident reporting and mandatory human verification of every output remain non-negotiable, whatever the class.
A procurement checklist
A checklist you can lift into a business case:
- Confirm the product is MHRA registered, and note the class and regime (UKCA or EU MDR).
- Read the certified intended purpose and check it matches how you intend to use the tool.
- If CE marked, confirm it is accepted for GB use until 30 June 2028 and ask the vendor's UKCA plan.
- Confirm completion of the DTAC assessment and inclusion on the NHS England AVT Supplier Registry.
- Confirm the clinical-safety documentation (DCB0129 from the vendor; plan your DCB0160).
- Confirm data handling, audio retention, and that patient data is not used to train models without explicit consent.
- Confirm the post-market surveillance and incident-reporting process.
- Confirm the workflow keeps a clinician reviewing and approving every output.
For the classes themselves, see our plain guide to medical device classification, and for the regime nuance, our piece on UKCA versus EU MDR Class IIa.
Frequently asked questions
What does Class IIa mean for a Trust buying an AI scribe? It means an independent body has certified the product for a higher-risk intended purpose, so part of the assurance is externally validated. Your local governance, clinical-safety and consent duties still apply.
Does Class IIa remove our local clinical-safety work? No. It lets your local safety case reference an external assessment, but you still need a Clinical Safety Officer, hazard log, safety case and DPIA. The responsibility for safe deployment stays with the Trust.
Who is liable if a certified scribe makes an error? Responsibility follows the boundary: the manufacturer for device conformity, the Trust for deployment and governance, and the clinician for verifying outputs. Using a tool outside its intended purpose shifts risk back to the deployer.
Is a Class IIa tool guaranteed to be accurate? No. Certification assesses process and intended purpose against a standard; it does not guarantee error-free outputs. Human verification and post-market surveillance remain essential.
What is the single most important thing to check? The certified intended purpose and scope, and that your intended use matches it, alongside MHRA registration. The headline class matters less than what exactly was certified.