Aipedia: Occupational health providers

AI in fitness-for-task and safety-critical medicals

Aipedia, by the Apex broking team · Last reviewed 2026-10-27

AI can support safety-critical medicals, but a competent clinician must make and sign the fitness decision. A false negative that clears an unfit driver or operator can injure third parties as well as the worker.

Part of: AI in occupational health providers

In short

OH providers carry out fitness-for-task medicals for drivers, rail staff, offshore workers and others in safety-critical roles. AI tools can score questionnaires, analyse test results and draft outcome certificates. The main risk is a false negative: a condition affecting alertness, vision or consciousness that the tool does not flag. Such decisions have significant effects on workers and engage UK GDPR Articles 22A to 22D if made solely by automated means. GMC para 7 requires balanced, accurate reports. Third-party injury claims after an OH clearance raise questions about how PI treats bodily injury.

How AI supports safety-critical medicals

Safety-critical medicals follow structured protocols: health questionnaires, vision and hearing tests, blood pressure, sometimes drug and alcohol testing, and a clinical interview. Providers may use AI to pre-score questionnaires, check results against protocol thresholds, flag inconsistencies between declarations and records, and generate a fit, fit with restrictions or unfit certificate for review. Large contracts for transport or energy clients can involve high volumes of medicals, so automation is attractive.

The weakness is that serious conditions can show up in what a worker does not say, or in subtle findings a protocol misses. A tool that checks answers against thresholds cannot notice evasive answers, a recent medication change or a pattern across several borderline results.

Employer clients may also put pressure on turnaround. A fit outcome lets a worker start or return quickly, while a referral for further assessment costs time and money. AI tools that make clearance faster can make it easier to skip the clinical interview that would have caught a problem.

The duties that apply

Report quality: the GMC's employment disclosure guidance para 7 requires reports not to be false or misleading and to give a balanced opinion within the doctor's direct knowledge.
Consent and scope: para 115 requires the worker to understand the purpose and consequences of the medical and limits disclosure to relevant information.
Software incidents: GMP para 73, as noted in the GMC's AI learning resource, requires adverse incidents involving software to be reported.
Automated decisions: fitness-for-work decisions with significant effects fall within UK GDPR Articles 22A to 22D if made solely by automated processing, requiring explicit consent or a contract or legal requirement plus Article 9(2)(g), with human intervention available.
Other registrants: OH nurses and technicians working under protocols remain accountable under NMC and HCPC standards for records and scope of practice.

What can go wrong

Illustrative scenario: a bus driver completes a medical questionnaire and does not mention recent episodes of daytime sleepiness. Blood pressure and vision are within protocol limits, and the AI tool issues a provisional fit outcome, which a busy clinician signs without an interview. Months later the driver falls asleep at the wheel and injures pedestrians. Claims follow against the employer, which asks whether the OH medical was adequate, and the OH provider's process is examined.

Other risks include tools that do not apply role-specific standards, automated outcomes sent directly to employers, and drift when protocols change but the software does not. Unfair unfit outcomes also cause harm: a worker wrongly restricted may lose earnings and bring a complaint.

How PI usually responds

A claim that a fitness-for-task medical was negligent is usually a professional indemnity matter for the OH provider and the signing clinician, subject to the policy terms. Third-party injury claims need care: some PI wordings exclude or limit bodily injury, while medical malpractice policies are designed to respond to it, so check which wording you have and whether it covers claims by third parties as well as the worker. Some policies, and one medical defence organisation's published position, assume the clinician made the decision. See bodily injury exclusions in PI and does OH indemnity cover AI mistakes.

Practical controls

Require a competent clinician to review and sign every safety-critical outcome.
Use role-specific standards and update the tool when they change.
Interview the worker where answers or results are borderline or inconsistent.
Never send automated outcomes direct to the employer.
Record the clinician's reasoning, not just the tool's score.
Audit a sample of fit outcomes for missed risk.

How Apex helps

Apex arranges PI for OH providers that carry out safety-critical medicals. A broker checks how your policy treats third-party bodily injury and AI-assisted outcomes; see OH professional indemnity.

Sources

Frequently asked

Can AI issue a fitness certificate for a safety-critical role?

It can draft one, but a competent clinician should review and sign it. A solely automated fitness decision with significant effects engages UK GDPR Articles 22A to 22D, and indemnity may assume a human made the clinical decision.

Can third parties claim after an OH clearance?

Third parties injured by a worker may claim against the employer, and the employer may look to the OH provider if the medical was inadequate. Whether the provider's policy responds depends on how it treats bodily injury and third-party claims.

What is a false negative in an OH medical?

A result that clears a worker as fit when they have a condition that makes them unsafe for the role. In safety-critical work this is the most serious error, because the harm can extend to colleagues and the public.

Should the clinician interview every safety-critical worker?

Not necessarily every one, but the protocol should require an interview where answers or results are borderline, inconsistent or new. A clinician signing outcomes should be able to show they considered the evidence, not just accepted the tool's score.

Check your indemnity covers how you use AI

Tell us about your practice and the work you do. A broker will review how your current cover treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start an indemnity quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.