AI in healthcare: Do you really know who is treating you and who is responsible when things go wrong?

Artificial intelligence is no longer just an academic debate for the future of healthcare in the UK.  Most people do not know that it is already being used within the NHS to support clinical decision-making and patient care on a daily basis.

Whilst the medical device industry and entrepreneurs continue a lucrative race to the market, the legislators and regulators in the UK have been unable to keep up with the rapid pace of change.  This has eroded patient’s rights.

The Medical Protection Society, a world leading organisation representing doctors, dentists and healthcare professions, has recently called on the government to address “the widening gulf” between AI and the law.  Their major concerns centre around their fear that they are in line to shoulder the legal responsibility if AI-assisted healthcare tech goes wrong whilst the companies behind the products remain “shielded”.

For injured patients, however, that raises an equally if not more important question namely if AI causes or contributes to avoidable harm leading to serious injuries or death, who becomes ultimately accountable to them?

Examples of how AI is currently being used in healthcare in the UK

The use of AI products in healthcare is on a steep rise.  A few examples of AI technologies currently being used in practice are set out below, but the list is now extensive:

• Symptom checkers such “Visiba” as the NHS 111 online service who are trialling AI to assist patient triage.
• Diagnostic tools such as “Google DeepMind” which has been implemented at Moorfield’s Eye Hospital to help identify serious eye conditions quickly and accurately. Another example is “AI” which has been rolled out to radiologists in over 40 NHS Trusts to identify chest conditions such as lung cancer.
• Predictive Analytics such as BloodImage! Which looks at blood samples to spot abnormalities.
• AI-enabled ambient scribing products to assist in transcribing medical appointments and summarising key medical information.

Who is liable when something goes wrong with AI?

Medical product liability claims in England and Wales are normally governed by the Consumer Protection Act 1987, which is now over 40 years old. The Act was drafted before the age of the internet when most products were manufactured in domestic factories and sold in high street shops.  The legislators in the 1980s understandably did not envisage and make provision for AI systems.

The Consumer Protection Act 1987 was introduced into UK law with the aim of ensuring that products supplied to the UK market are safe and that those who supply defective products that cause harm can be held legally accountable. Unlike a typical claim for medical negligence, the Act is supposed to be a strict liability regime, meaning that there is no need to prove negligence, only that the product was ‘defective’ and that it caused the damage complained of.   Whilst each case turns on its own facts, in reality it can be extremely difficult for injured patients to bring viable product liability claims against big tech companies, many of whom are based abroad.   On the most basic level, some may argue that AI software, for example, is not a “product” for the purposes of the Act.

Historically, the position for bringing clinical negligence claims has been relatively clear. If a patient suffers avoidable harm, the law asks whether a clinician acted negligently and whether that caused the injury.  The use of AI in healthcare complicates that analysis. A clinician may now be relying, sometimes heavily, on a recommendation generated by a software tool. That recommendation may feel authoritative, particularly where the tool is widely used or endorsed within the NHS. But if the output is wrong due to a product defect or malfunction, the chain of responsibility becomes much harder to untangle.  Who is responsible for training the clinicians in the safe use of the AI tech? The list of possibilities is endless. The same concern works in both directions. If a clinician follows an AI recommendation that proves to be wrong, they may face criticism for relying on it. If they reject an AI output and the patient is later harmed, they may also be asked why they departed from it. That leaves doctors in a difficult position, but it also leaves injured patients facing uncertainty about where responsibility truly lies.

The Medicines and Healthcare Regulatory Authority (MHRA) who regulate the safety of medical devices have been unable to solve the problem so far.  AI systems used in healthcare are typically regulated as medical devices. They must meet safety and performance standards before being deployed. But that framework is primarily concerned with whether a product can be used, not who is responsible when its use causes harm. The Medical Protection Society’s warning is important because it shows that regulation and liability are not the same thing. A product may be approved for use, but that does not automatically give an injured patient a clear answer as to who should be held accountable if the technology contributes to avoidable harm.

The risk of patients being caught in the middle

For injured patients, this lack of clarity is more than an academic concern. It risks more complex claims, longer investigations, and greater delay. There is also a more fundamental concern: if responsibility is sufficiently diffuse, patients may find themselves without a clear route to accountability.

An issue which seems to be continually overlooked in this debate is that of informed consent. Patients are entitled to understand the material risks associated with their treatment.  In a world where AI is influencing clinical decision-making, that raises important questions about transparency. Are patients being told that AI tools are being used in their care? Do they understand the extent to which a decision has been shaped by software? Are the limitations including the risk of error or bias being explained?  A failure to do so could expose clinicians to liability.

Conclusion

AI undoubtedly has the potential to improve care. But it also introduces new risks. The Medical Protection Society is quite right to identify a liability gap between the law and the pace of technological change. From a claimant perspective, the priority must remain unchanged: patients who suffer avoidable harm should have access to answers, accountability, and fair compensation. That means ensuring that clinicians, NHS bodies, regulators, and technology developers each have clearly defined responsibilities when AI is used in healthcare. Without clearly defined responsibilities, injured patients may face unnecessary complexity, delay, and dispute when trying to understand how an AI-assisted error happened.

How we can help?

If you or a loved one has been harmed in circumstances where AI or technology may have played a role, our leading clinical negligence team can help investigate what happened and advise on your options.

About the authors

Tom Jervis is a Partner at Anthony Gold Solicitors LLP. He specialises in high-value clinical negligence and personal injury claims, acting for individuals who have suffered serious and life-changing injuries. He is a product liability specialist and has successfully resolved claims involving robotic surgery and is currently advising patients in cutting-edge claims where AI programmes may have played a part in how they have been harmed during treatment.

Theo Libman is a Trainee Solicitor at Anthony Gold Solicitors LLP. He joined the Injury and Medical Claims Team in September 2025 and is assisting Jackie Spinks, Dr James Piers, Alex Stutt & Tom Jervis with a number of their clients’ cases. Theo graduated from the University of Birmingham in 2022 with a First Class in his Law Degree and is currently awaiting the results of his SQE 2 Examinations.  Theo  has a wide range of experience in client facing roles and brings the benefit of this expertise to his day to day activities.