Who carries responsibility for NHS AI errors? Concerns over developer accountability

Accountability under scrutiny

According to Digital Health, Randeep Sidu, a non-executive director at the NHS Business Services Authority, questioned whether the government’s response to the national healthcare AI commission would provide sufficient accountability for developers. Speaking at BMJ Future Health 2026, he reportedly warned that clinicians could be left responsible when technology makes errors.

The concern reported by Digital Health is about how responsibility is distributed, rather than simply whether AI can be introduced into healthcare. For surgeons considering the organisational implications of AI, that distinction may be important: responsibility for a technology’s performance and responsibility for its use are related questions, but should not be treated as interchangeable in this debate.

Across the AI lifecycle

Digital Health reported that Sidu raised concerns about responsibility across the AI lifecycle. This places the discussion beyond the individual moment when a clinician encounters an AI output, asking how accountability is understood across the technology’s development and use.

The supplied Digital Health account does not set out a detailed allocation of responsibilities between developers, NHS organisations and clinicians. It is therefore not yet clear how the concerns Sidu raised would translate into specific arrangements, or which aspects of responsibility the government’s eventual implementation work will address. His reported warning should not be read as an established finding that clinicians will necessarily carry responsibility for every technology error.

Safe operating conditions

According to Digital Health, safe operating conditions were another part of Sidu’s concerns. Alongside accountability across the lifecycle, this raises a question about the circumstances in which an AI system is expected to function safely, and how those circumstances relate to responsibility if something goes wrong.

It is not yet clear from the supplied Digital Health account what particular operating conditions Sidu had in mind, or how they would be described in practice. The report’s summary does not provide a technical framework, performance thresholds or examples involving surgical services. Any application of this discussion to a particular surgical technology would therefore go beyond the information supplied.

Contracts and developers

Digital Health also reported concerns about contractual accountability. In the discussion as summarised, this sits alongside Sidu’s question about whether the government’s response would make developers sufficiently accountable, and his warning about clinicians potentially being left responsible for technology errors.

The supplied Digital Health account does not describe proposed contractual terms or explain how any future arrangements would operate. It is not yet clear whether the forthcoming implementation detail will resolve the concerns reported, or how responsibilities between the different parties would be expressed. The distinction between a reported concern and a settled policy position remains important.

What remains to be clarified

According to Digital Health, a detailed government implementation roadmap is expected by spring 2027. That expected timetable provides a point at which further detail may emerge, but the supplied account does not establish what the roadmap will contain.

For a surgical audience, the debate may be useful as a way of framing questions about shared responsibility, without assuming that those questions have already been answered. Digital Health’s report presents Sidu’s concerns about developer accountability, safe operating conditions and contracts; it does not establish a final allocation of responsibility for NHS AI errors.

This article is an overview for information only; readers should consult the original sources and appropriate professional guidance.