Хиймэл оюун ухааны системүүд нэвтрэхийн хэрээр NHS-д өгөгдөл, хандалт, хариуцлагыг тасралтгүй хянах шаардлага нэмэгдэж байна.
Их Британийн Эрүүл мэнд, нийгмийн халамжийн яам NHS-ийн Холбооны өгөгдлийн платформ (FDP)-ын гэрээг хугацаанаас өмнө цуцлах заалт хэрэгжихээс өмнө дахин нягтална гэж мэдэгджээ. Үндэсний дэд бүтэц, үйлчилгээний өөрчлөлтийн газраас төслийн нийт өртгийг 1.1 тэрбум фунт, хүлээгдэж буй өгөөжийг 808 сая фунтээр шинэчлэн тооцсон нь бизнесийн үндэслэлд тавих хяналтыг нэмэгдүүлсэн. Гэвч нийлүүлэгчийг солих нь өгөгдлийн засаглалын суурь асуудлыг өөрчлөхгүй.
FDP нь төвлөрсөн нэг платформ бус, эмнэлгийн трастууд болон нэгдсэн тусламж үйлчилгээний зөвлөлүүдийн орон нутгийн систем, NHS England-ийн үндэсний системээс бүрдэх холбоост бүтэцтэй. Энэ загвар өгөгдлийн хяналтыг холбогдох байгууллагуудад үлдээдэг ч систем хоорондын техникийн хамаарлыг тогтмол зураглаж, эрсдэлийг үнэлэх шаардлага үүсгэдэг. NHS England-ийн мэдээлснээр үйлчилгээ үзүүлэгчдийн дөрөвний гурав нь хуримтлагдсан техникийн өрөө шийдвэрлэх хөрөнгө оруулалт шаардлагатай гэж үзсэн бөгөөд ажилтнууд нэг ажлыг гүйцэтгэхдээ олон систем ашигладаг хэвээр байна.
2026 оны тавдугаар сарын эцсийн байдлаар 139 траст, 35 нэгдсэн тусламж үйлчилгээний зөвлөл FDP-д холбогдсон байв. Хиймэл оюун ухааны систем, агентууд аппликейшн, өгөгдлийн сан болон гуравдагч талын үйлчилгээнд холбогдох тусам хэн эзэмшдэг, юунд хандаж ямар үйлдэл хийх эрхтэйг мэдэх шаардлага нэмэгдэнэ. Загварын хандалт эсвэл зан төлөв ажиллах явцдаа өөрчлөгдөж болох тул тогтмол хугацааны үнэлгээ дангаараа хангалтгүй; эрсдэл нь зөвлөмж буруу гарахаас гадна систем юунд нэвтэрсэн, юу өөрчлөгдсөн, үр дүнд хэн хариуцахыг цаг тухайд нь тогтоож чадахгүйд оршино.
Ирээдүйн платформ нь хиймэл оюун ухааны систем, өгөгдлийн эх үүсвэр, эзэмшигч болон зөвшөөрлийн бүртгэл хөтөлж, загвар, өгөгдөл эсвэл эрх өөрчлөгдөх бүрд эрсдэлийг дахин үнэлэх боломжтой байх ёстой. Мөн хэрэглээний эрсдэлд тохирсон хяналт тавьж, бодит цагийн аудитын мөр гарган, өрөөсгөл байдал болон өөрчлөлтийг илрүүлэх шаардлагатай. Өгөгдөл орон нутгийн системээс үндэсний системд дамжихад хуулийн хариуцагч хэн байх, инцидентэд хэн хариуцах, хандалтын зөвшөөрлийг хэн баталсныг тухайн шийдвэр, эрхтэй холбож бүртгэх нь засаглалыг хангах гол нөхцөл юм.
Дэлгэрэнгүйг эх сурвалжаас харах
↓Эх сурвалжийг нээх ↓
Picture February 2027. The government triggers the break clause, Palantir packs up, a new supplier arrives with a better price and a British postcode, and the migration begins.
What, precisely, would change for the health service? The federated and ownership at each level keeps responsibility and control in the hands of the data controllers.
The pressure to reconsider the Federated Data Platform is, however, understandable. The National Infrastructure and Service Transformation Authority has revised the program’s projected lifetime cost to £1.1 billion, against projected benefits of £808 million.
The widening gap between cost and projected benefits has intensified scrutiny of the program’s business case.
The Department of Health and Social Care has confirmed that the contract will be reviewed before the break clause, and questions about value, sovereignty and supplier dependency deserve serious attention.
But changing suppliers will not address the deeper governance problem.
The federation is the point and the challenge
The FDP was deliberately designed as a federation. Trusts and integrated care boards operate local instances, while NHS England operates the national instance. Keeping data local by design creates multiple layers of ownership, which can strengthen accountability by keeping control with the relevant data controllers.
But that same federated model creates technical dependencies across systems and organizations that need to be continuously mapped, risk-assessed and managed.
We are not talking about a single platform here. It is a network of interconnected platforms sitting within an estate that was already complex. NHS England says three-quarters of providers have identified investments needed to address their technical debt. Staff still move between multiple systems to complete a single task, while data sharing across organizational boundaries remains inconsistent.
Scrap the FDP and hand the work to a British supplier, and the underlying conditions do not change. The data controllers remain. The legacy systems remain. The technical dependencies remain. So do assurance mechanisms designed for a far more static technology environment.
The NHS has no shortage of governance roles and frameworks e.g., the Data Security and Protection Toolkit, the Cyber Assessment Framework, data protection impact assessments, senior information risk owners, Caldicott Guardians and the Digital Technology Assessment Criteria. The list goes on. They are important safeguards. However, they largely assume that an organization can assess a system at a point in time and expect the resulting picture to remain reliable.
In the modern era of AI-native data federation, that assumption is becoming untenable.
AI turns a governance gap into an operational risk
As AI systems and agents connect to existing applications, datasets and third parties, the governance surface expands past the data platform itself.
Security teams need visibility into more than just the infrastructure underneath. They need to know who owns each AI system and what it is permitted to reach and do.
This is not a future problem. By the end of May 2026, 139 trusts and 35 integrated care boards were live on the FDP. The governance question already spans a substantial network of local and national instances, not one platform or one supplier.
Existing AI assurance approaches are not designed for this level of dynamism. The MHRA’s approach to AI as a medical device and NHS England’s AI governance guidance are valuable foundations. But periodic assessment cannot by itself govern an agent whose access or behavior can change at runtime.
The risk is not simply that an AI system makes a poor recommendation. It is that nobody can establish quickly enough what the system could access, what changed or who is accountable for the outcome.
Judge the next platform on what it can prove today
Any future NHS data and AI platform should be judged on more than cost or sovereignty. The test is whether it can provide continuous visibility and accountability across the wider estate.
In practical terms, it should be able to:
– identify the AI systems operating across the estate and maintain a usable inventory of the models, data sources, owners and approvals involved
– reassess risk when models, data or permissions change, rather than waiting for an annual return or scheduled review
– apply controls appropriate to the use case, recognizing that a discharge-scheduling tool and a cancer-pathway tool do not carry the same risk
– produce evidence and audit trails in real time, making compliance a query rather than a fire drill
– detect drift and bias, while routing approvals and incidents to named individuals who understand that accountability is theirs
Accountability must also survive an incident. Who is the legal controller under UK GDPR when a trust’s data passes through a national instance? Who owns an incident that begins locally and ends nationally? Who approved a particular access grant, model or action?
Those answers should be recorded within the operating system of the organization, connected to the decisions and permissions they describe. They should not have to be reconstructed from documents written months earlier.
A new supplier is not the solution
The NHS is not the only organization facing this. Water, energy, transport and local government are all building connected estates with layered ownership and suppliers they cannot always fully see.
The break clause may change the supplier for the NHS but it will not fix the governance.
A real solution would enable the NHS to identify what data is running across the estate, what systems can access it, who is accountable for data protection controls at each level, and what risks are emerging as the system continues to evolve. If it can’t, changing the name on the contract is not a solution, it is just a distraction.
We’ve reviewed and ranked the best small business software.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

