Health

Doctors and the NHS could be sued for mistakes made by AI tools, report warns

Guardian Health2 h ago
An empty hospital workstation with a computer screen
An empty hospital workstation with a computer screenPhoto: Tima Miroshnichenko / Pexels

Medical Protection Society (MPS), one of the UK's largest medical liability protection bodies, has published a new report saying that AI tools used increasingly in clinical settings could put doctors and NHS services at legal risk. According to the Guardian, the report warns that existing indemnity frameworks are not keeping pace with the spread and speed of AI in the clinic.

Clinical AI tools span a wide range, from imaging software and clinical decision support systems to symptom checkers and triage bots. NHS pilots are running in dozens of trusts, particularly in radiology and general practice. MPS says that when these systems make incorrect recommendations, 'who is liable' is currently not well defined.

The report defines three main liability scenarios. The first is cases where the clinician accepts the AI recommendation without challenge; the second is where the clinician dismisses the AI and misses a correct answer; the third is errors arising from internal faults in the AI system. Insurance and litigation work differently in each.

MPS chief executive Stephen Ledger told the Guardian: 'Doctors are currently protected against judgement errors, but in cases involving AI it is difficult to separate which decisions came from the human and which from the machine. That creates a significant gap in existing indemnity policies.'

There is also uncertainty over manufacturer liability. Most AI software vendors state in their contracts that the system is 'a support tool, not a clinical decision-maker.' That phrase makes it easier in litigation to shift responsibility to the clinician. According to the Guardian, the report argues this contractual structure needs to be revisited.

The report suggests NHS managers are also exposed. Deploying an AI system in clinical workflows without adequate validation could draw trust executives into the chain of liability. MPS recommends that procurement processes require clinical validation and risk assessment steps as standard.

On the regulatory side, the MHRA (the Medicines and Healthcare products Regulatory Agency) has said it is preparing new guidance for AI-based medical software. The guidance is expected to be completed this year and should clarify approval pathways for clinical use. The European Union's AI Act also adds further obligations for medical AI systems.

Patient advocacy groups stressed the report's significance for patient safety. A Patient Safety Learning spokesperson said: 'Patients want to know who they can hold to account when AI gets it wrong. Uncertainty undermines trust and slows the health system's ability to adapt.'

Changes in everyday clinical practice are expected. MPS recommends that clinicians develop the habit of documenting the reasoning behind accepting or rejecting an AI recommendation. Logging AI output and clinical decision rationale separately in patient notes will be critical for both defence and accountability in future.

The Guardian writes that the report calls for a new 'liability-sharing compact' between the NHS, AI vendors and clinicians. While parallel debates are running in many countries, the report urges the UK to move insurance and liability reform at the same pace as the spread of clinical AI. This is not medical or legal advice.

This article is an AI-curated summary based on Guardian Health. The illustration is a stock photo by Tima Miroshnichenko from Pexels.

Read next