Aipedia: Doctors and medical practices

AI in medical practice: rules, risks and PI

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

UK doctors can use AI in clinical work, but the GMC holds them responsible for their own decisions and records whatever tool helped. That matters because NHS schemes, defence organisations and insurers may assume a human stayed in control.

Part of: AI in doctors and medical practices

In short

The GMC has no standalone AI standard: Good medical practice 2024 applies to AI as it does to any other tool, including competence (paras 3 and 13), records (paras 69 to 71) and incident reporting (para 73). NHS England has detailed rules for AI scribes, and on 6 October 2026 the government accepted all 44 recommendations of the National Commission on AI in healthcare. A licensed doctor must hold appropriate indemnity under the Medical Act 1983 s44C. NHS schemes cover NHS clinical negligence but not private work, regulatory hearings or data breaches, and one defence organisation's discretionary help with AI depends on a human remaining the decision maker.

How UK doctors are using AI

AI is moving from pilot projects into everyday medicine. The GMC commissioned research after an Alan Turing Institute survey in which 29% of nearly 1,000 doctors said they had used AI in the past year. Its report, Understanding doctors' experiences of using AI, found efficiency gains, although some doctors spent extra time checking the output of decision-support tools. In the GMC's Workplace experiences 2026 research (13 August 2026), AI and better information systems were among the most cited improvements in doctors' working lives.

The main uses in UK practice are:

ambient voice technology that listens to a consultation and drafts the note and letter;
diagnostic decision support and risk prediction;
image triage and second reads in radiology;
online consultation and care-navigation tools that sort patient requests in general practice;
automated filing of lab results and hospital correspondence;
general-purpose generative AI chatbots for drug queries, guideline summaries and drafting.

Doctors in the GMC research named biased data, over-reliance and deskilling as the main risks. Generative AI users said they avoided entering confidential patient data.

The rules: who says what

No single body regulates AI in medicine. The GMC regulates doctors but says it does not approve technologies; the MHRA regulates software and AI as medical devices. The GMC's AI learning resource applies existing Good medical practice principles: use professional judgement, stay within your competence, discuss AI use with patients, report adverse incidents involving software and digital tools (para 73), and raise concerns if a technology may put patients at risk.

NHS England's ambient scribing guidance sets the most detailed operational rules, including MHRA registration, clinical safety standards DCB0129 and DCB0160, a DPIA and telling patients. The National Commission into the Regulation of AI in Healthcare reported on 10 September 2026, and the government response of 6 October 2026 accepted all 44 recommendations. On 17 September 2026 the GMC and other regulators signed a joint statement committing to shared UK-wide AI principles, with no timetable yet. Our guide to the GMC, NHS England and MHRA guidance separates binding duties from advice.

The main risks for doctors

Automation bias. A clinician accepts an AI suggestion, or is falsely reassured by a missing flag, and a diagnosis is delayed.
Record errors. A scribe mishears a drug, dose or side, and the error is copied into letters and onward care. GMP para 69 still requires records to be clear, accurate and contemporaneous.
Under-triage. An online tool ranks an urgent request as routine.
Confidentiality. Patient data entered into a public chatbot, contrary to GMC confidentiality guidance and data protection law.
Misleading documents. Fit notes, fitness letters or expert reports signed without the assessment or checking that Domain 4 requires.
Unclear liability. NHS England calls AI liability "complex and largely uncharted" and warns it may default to the trust or practice.

The indemnity position

Under Medical Act 1983 s44C, a licensed doctor who practises must have "an indemnity arrangement which provides appropriate cover", and the GMC can refuse or withdraw a licence without it. No minimum limit is set, but GMP Domain 4 says cover must reflect the full scope of practice and be kept under regular review. See doctors and indemnity: the GMC requirements.

NHS work is usually covered by NHS Resolution's state schemes: CNST for trusts and CNSGP for NHS general practice from 1 April 2019. These do not cover private work, occupational health, regulatory and fitness-to-practise matters, or data protection breaches. Doctors fill those gaps through a medical defence organisation, whose help is discretionary, or an insured medical malpractice policy. One defence organisation's published position offers assistance with AI only where a human remains the decision maker, and excludes fully autonomous systems, product liability, system defects and data breaches. Private GPs can read our private GP indemnity guide and CNSGP and private work explainer.

What to watch in 2026 and 2027

Under Recommendation 24 of the government response, a working group of DHSC, the MHRA, CQC, NHS Resolution, the PSA and the professional regulators will clarify professional responsibilities in AI use, but "will not determine or redistribute responsibilities". NHS England and NHS Resolution will continue work on indemnity guidance. The MHRA plans draft guidance on AI device change management by December 2026, a classification consultation in 2027 and an implementation roadmap by spring 2027. The government has not committed to product liability reform, which some defence organisations have called for.

Aipedia guides for doctors

GMC, NHS England and MHRA AI guidance explained: what is binding and what is advice.
AI scribes in consultations: NHS England requirements and record-keeping duties.
AI diagnostic decision support: automation bias and who is liable for a missed diagnosis.
AI in radiology and image reading: agreeing or disagreeing with an AI flag.
Generative AI chatbots for clinical questions: hallucinations, confidentiality and competence.
AI-drafted clinic letters and referrals: errors that travel to the next clinician.
AI triage and online consultation in GP: under-triage and the practice's duty.
AI fit notes and fit-to-fly letters: why an adequate assessment is still required.
AI in prescribing and medication safety: alert fatigue and overridden warnings.
AI for lab results and correspondence: abnormal results that are never actioned.
AI in medico-legal expert reports: disclosure to instructing solicitors.
Patient chatbots for private clinics: when a chatbot gives clinical advice.
AI-generated patient complaints: responding without admitting liability.
Does medical indemnity cover AI mistakes?: conditions and exclusions.
CNST, CNSGP and AI: where NHS indemnity stops.

How Apex helps

Apex arranges medical malpractice and professional indemnity cover for doctors and clinics, particularly for private, occupational and medico-legal work outside NHS schemes. A broker reviews your cover and how your policy treats AI-assisted work; see also does PI cover AI mistakes.

Sources

Frequently asked

Does the GMC allow doctors to use AI?

Yes. The GMC has no standalone AI standard and does not approve technologies. It expects doctors to apply Good medical practice: use professional judgement, work within their competence, keep accurate records, discuss AI use with patients and report adverse incidents involving software or digital tools under para 73.

Is indemnity compulsory for UK doctors?

Yes. Section 44C of the Medical Act 1983 requires a licensed doctor who practises to have an indemnity arrangement providing appropriate cover. It can be insurance, an indemnity or both. No minimum limit is set, but cover must reflect your full scope of practice, including any AI-assisted private work.

Does NHS indemnity cover mistakes made with AI tools?

For NHS clinical work, CNST and CNSGP usually respond to clinical negligence claims whatever tools were used, subject to scheme rules. They do not cover private work, occupational health, regulatory hearings or data protection breaches, and the CNSGP scope table says nothing specific about AI or software suppliers.

Who is liable if an AI tool gets it wrong?

It is unsettled. The GMC says clinicians keep responsibility for decisions, while developers and employers stay responsible for their parts. NHS England warns liability may default to the trust or practice if no supplier accepts it. A government working group will clarify responsibilities but will not redistribute them.

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.