-
Healthcare
-

When The Health Service Learns to Defend Itself

By
Distilled Post Editorial Team

A trust in south London signs a contract worth just under £380,000 to bring artificial intelligence into its consulting rooms, a sum that on its own barely registers against an NHS budget measured in the hundreds of billions. Multiply it across the roughly 120 acute trusts in England and the number becomes harder to dismiss, edging towards £45 million, much of it justified less by clinical benefit than by the evidentiary value of a transcript, timestamped and complete, that can be produced if a patient later claims something went wrong. Ambient voice technology was sold to the service as administrative relief, freeing clinicians from the tyranny of typing so they might look patients in the eye again. That case still holds. But a second, quieter justification has crept alongside it, and it says something less comfortable about where the relationship between the NHS and the people who use it now stands.

Some trusts are asking to retain two years of audio and transcript data, a retention period that has little to do with clinical utility and everything to do with the statute of limitations on negligence claims. NHS Resolution, the body that manages litigation risk across the service, concedes that ambient scribes are producing more consistent clinical records, while stopping short of saying what that consistency will do to the volume or cost of claims. That hesitation is itself informative. An institution does not usually build defensive infrastructure this expensive while remaining agnostic about whether it needs defending.

It needs defending, in part, because patients have found their own version of the same tool. Generative AI has lowered the barrier to writing a detailed, procedurally literate complaint, and complaints teams already thinned by years of cost pressure are struggling to process what arrives. Board papers from Liverpool's hospital group describe submissions that are technically dense and difficult to parse. A chatbot can now produce in minutes the kind of document that once required a solicitor or a confident advocate, regardless of whether the underlying grievance has grown any more serious. Medical negligence specialists report a related problem, with AI tools regularly overstating the strength of a claim, misreading clinical detail, or treating a trust's routine expression of regret as an admission of fault. The result is a complaints system absorbing a wave of submissions that are simultaneously more sophisticated and more likely to be mistaken.

What makes this worth attention beyond the technology itself is what it implies about trust as an operating principle inside the health service. AVT exists because a trust no longer assumes a dispute will be resolved through conversation and professional judgement; it wants a recording. Patient-drafted, AI-assisted complaints exist for a mirrored reason, because many people no longer believe an informal word with PALS or a ward sister will be taken seriously without a paper trail behind it. Each side is reaching for the tool available to it, reasonably enough on its own terms. A service that increasingly prepares for bad faith on the other side of the desk is describing its own erosion even as it tries to protect against it.

This ought to concern NHS leadership and the Department of Health beyond questions of cost, though the cost is real at a moment when trusts are being told to find productivity savings elsewhere. Audio retention practices are developing trust by trust, with no evident national standard governing how long recordings are kept, who can request them, or how they interact with subject access requests. That is a governance gap waiting to become a headline, most likely the first time a family discovers that a two-year-old recording of a difficult consultation has been kept, for reasons never made explicit, and used against their case rather than in support of it. If ambient technology is to remain a tool for better care rather than a quiet instrument of institutional self-protection, someone in government will need to decide, deliberately, what its purpose is meant to be. At present that decision is being made by default, contract by contract, trust by trust, and the answer taking shape is not one most patients would recognise as care.