As Liberal Democrats, we believe in universal access, clinical autonomy, local accountability, and innovation that serves the patient rather than the platform. Labour’s 10-year plan for the health service threatens each of these foundations. It shifts decision-making power from clinicians to digital triage apps, replaces continuity of care with walk-in hubs, and centralises England’s patient data under the control of Palantir, a US surveillance firm with no democratic oversight. The plan anticipates fewer staff, conditions access on risk scores and outcomes, and introduces no new safeguards on how patient data is routed, monetised, or reused. This is not modernisation. It is a quiet, systemic repurposing of the NHS, and the public deserves to understand the full implications.
Digital Gatekeeping
At the centre of Labour’s digital vision for the NHS is a radical shift in how patients access care. By 2028, the NHS App will become the universal entry point to NHS care in England (Labour 2025: 10). This includes triage, appointment booking, and condition management, all of which are presently core functions of local practices and NHS staff. The plan outlines a “My NHS GP” feature to route all access digitally (Labour 2025: 11, 31).
Yet placing digital triage at the heart of NHS access introduces serious and well-documented clinical risks. AI symptom checkers show error rates of 20–40% depending on symptom complexity, often under-triaging serious cases or giving false reassurance (Fraser et al. 2022; BMJ 2020). They lack clinical context, are not accountable, and disproportionately fail older adults, patients with cognitive or language barriers, and those with multimorbidity (King’s Fund 2022).
Institutionalised Staffing Shortages
These risks can, of course, be mediated through medically professional oversight, a practice common on the continent, where digitalization is introduced to augment, rather than replace, medical professionals. However, rather than follow European best practice, Labour’s plan appears to institutionalise staff shortages as necessary to the functioning of the new digital NHS. By forecasting that ‘fewer staff than projected’ will be needed by 2035 due to anticipated efficiencies from automation, AI scribes, and redesigned roles (Labour 2025: 74), the plan builds systemic understaffing into the future model of care.
Cutting roles on the assumption of seamless substitution rarely works in complex systems like healthcare. Evidence from NHS digital implementation reviews shows that automation and role redesign frequently fail to deliver efficiency in practice due to clinical interdependencies and the unpredictable nature of care pathways (King’s Fund 2021; Health Foundation 2020). Rather than reducing labour, substitution redistributes pressure, deepens burnout, and increases the likelihood of unsafe gaps in safety-net care (GMC 2022; BMJ 2022). Digital tools can assist, but they cannot replace the presence, judgement, or adaptability of a trained clinical team operating under pressure (WHO and OECD 2020).
The End of Outpatients
As is to be expected from radical reductions in workforce expectations, Labour’s plan includes a restructuring of service delivery. By 2035, outpatient departments in England will be eliminated and replaced by “Neighbourhood Health Centres” responsible for diagnostics, monitoring, and follow-up (Labour 2025: 35). These are framed as flexible and multi-skilled, but there is no provision for clinical continuity, responsibility, or long-term therapeutic relationships.
However, removing that continuity risks far more than administrative confusion. Patients without a consistent clinical anchor are more likely to fall through gaps, face delays in diagnosis, and suffer from contradictory advice (King’s Fund 2018; Royal College of General Practitioners 2020). Complex or chronic cases (the very patients who use outpatient services most) depend on long-term therapeutic relationships (National Voices 2022). Labour’s plan dismantles that structure without offering an alternative, making the system more efficient for providers but more opaque and fragile for patients.
Disenfranchisement Risks
This structural overhaul marks, of course, a deeper shift in how care itself is conceptualised, away from relationships and clinical judgement towards predictive modelling and performance metrics. According to Labour, care pathways will increasingly be shaped by “value-based reimbursement,” patient-scored outcomes, and population-level “risk stratification” models to predict future healthcare needs (Labour 2025: 15, 52).
These metrics are intended to drive funding and routing but they risk replacing professional judgement with algorithmic scoring. Evidence from the US shows how such systems can go dangerously wrong. A 2019 Science study found that a widely used hospital algorithm systematically underestimated the health needs of black patients because it used previous healthcare spending as a proxy for medical need — despite the fact that Black patients often receive less care for the same conditions (Obermeyer et al. 2019). In the NHS context, similar logic could deprioritise patients with complex, chronic, or mental health conditions that do not fit the model well — or those who struggle to navigate digital platforms, including older adults, people with disabilities, and those facing language or literacy barriers (NHS England 2023; Digital Health News 2024).
You are the Product
What this means is that while the plan repeatedly refers to the NHS as a “learning system” that forecasts patient flows and allocates care based on live, data-fed feedback loops (Labour 2025: 48), it actually redefines the NHS as a national data infrastructure instead of a universal health service. In short, care is no longer something delivered on the basis of need alone, but allocated according to predicted value and system capacity.
Worse still, the system being built is not just predictive, it is extractive. NHS England already sells access to large patient datasets through its Secondary Uses Service and the General Practice Data for Planning and Research (GPDPR) programme, including to pharmaceutical companies and analytics providers (OpenDemocracy 2021; BMJ 2023). Under the Federated Data Platform, now run by Palantir, these flows could become fully systematised at national scale, without new oversight or opt-outs.
Palantir, a US firm founded with CIA seed funding through In-Q-Tel, was built to serve intelligence and law enforcement agencies. Its co-founder, Peter Thiel, is a prominent libertarian and political donor who has openly questioned the compatibility of democracy with freedom (Thiel 2009), and has invested heavily in technologies of control — from predictive policing to border enforcement (The Atlantic 2020; Wired 2020). Palantir’s core business has always been surveillance, behavioural analytics, and population-level prediction — not healthcare (O’Neil 2016).
So why does the company now hold a £330 million contract to coordinate NHS England’s care data infrastructure via the Federated Data Platform (Financial Times 2023)? Especially as the UK has no specific statutory safeguards preventing such data from being reused for non-clinical purposes or accessed by international actors (ICO, 2021; NHS England, 2023)?
The answer is that behind the scenes, NHS patient data itself is the incentive being used to attract investment. Indeed, our sensitive health data is the hidden asset, the sweetener that makes long-term private control profitable. And despite assurances of privacy, these data-sharing practices signal the breakdown of clinical confidentiality. The UK hosts one of the richest, most longitudinal health datasets in the world. With sustained access, Palantir gains a commercial foothold in one of the few datasets capable of training global-scale health AI (BMJ 2023, Guardian 2023).
Data-sharing initiatives like GPDPR already extract detailed records from GP practices for secondary use without meaningful transparency or public opt-in (OpenDemocracy, 2021; BMJ, 2023). Crucially, Although the data is pseudonymised, it is not truly anonymous. Patients can often be re-identified when linked with other datasets, especially given the granularity of medical records (OpenDemocracy, 2021; ICO, 2021).
This deep and big data operating in a legal vacuum enables broad secondary use and long-term commercial access, not just by Palantir but potentially other firms headquartered abroad. International cases reveal the risks: in the US, Medicaid data has been shared with ICE (Guardian, 2025); Palantir’s pandemic tools have raised privacy alarms (Wired, 2020); and Google’s Project Nightingale gave the company access to millions of patient records from Ascension without informing patients, using the data to develop AI tools under HIPAA’s “business associate” rules — entirely legal, but done without public awareness or consent (Copeland & Mathews, Wall Street Journal, 2019). In short, Labour’s plan kills privacy, sweeps aside consent, and your data could potentially be bought and sold, and used against you, without your knowing, or be able to find out.
Conclusion
We are a party committed to safeguarding the NHS as a public service; individual rights, and public services run in the public interest. Labour’s plan is not digitisation in the public interest. It is a structural transformation of the NHS into a predictive and commercial platform — one where access is modelled, outcomes are scored, and patient data becomes the commodity.
If the public is to retain confidence in the NHS as a universal service, then clinical oversight, data protections, and democratic accountability must come first. Unfortunately, it appears that it won’t.
Bibliography
BMJ, ‘NHS Federated Data Platform: Concerns over Palantir’s Role Remain’ (2023) BMJ https://www.bmj.com/content/381/bmj.p797 .
Copeland R and Mathews AW, ‘Google’s “Project Nightingale” Gathers Personal Health Data on Millions of Americans’ The Wall Street Journal (11 November 2019) https://www.wsj.com/articles/google-s-secret-project-nightingale-gathers-personal-health-data-on-millions-of-americans-11573496790 .
Financial Times, ‘Palantir Wins £330mn NHS Data Contract Despite Privacy Concerns’ (21 November 2023) https://www.ft.com/content/9f3b6d2c-00c6-44b5-81d0-4efbdf7e4f35 .
Fraser H and others, ‘Safety of Patient-Facing Digital Symptom Checkers’ (2022) 369 BMJ https://www.bmj.com/content/376/bmj.n1297 .
General Medical Council, State of Medical Education and Practice in the UK (2022) https://www.gmc-uk.org/about/what-we-do-and-why/data-and-research/the-state-of-medical-education-and-practice-in-the-uk .
Health Foundation, Shifting the Balance: Risks and Opportunities of Automation in General Practice (2020) https://www.health.org.uk/publications/reports/shifting-the-balance .
Information Commissioner’s Office (ICO), Anonymisation, Pseudonymisation and Privacy Enhancing Technologies (2021) https://ico.org.uk/media/about-the-ico/documents/4018606/anonymisation-intro-and-first-chapter.pdf .
King’s Fund, Technology and Innovation in the NHS (2021) https://www.kingsfund.org.uk/publications/technology-innovation-nhs .
King’s Fund, Understanding Quality in General Practice (2018) https://www.kingsfund.org.uk/publications/quality-general-practice .
Labour Party, Labour’s Plan to Modernise the NHS: A 10-Year Commitment to Health and Care (2025).
National Voices, Unlocking the Value of People with Lived Experience in Service Design (2022) https://www.nationalvoices.org.uk/publications/our-publications/unlocking-value-people-lived-experience-service-design .
NHS England, Federated Data Platform Procurement Notice (2023) https://www.england.nhs.uk/digitaltechnology/digital-data/fdp/.
Obermeyer Z and others, ‘Dissecting Racial Bias in an Algorithm Used to Manage the Health of Populations’ (2019) 366 Science 447.
O’Neil C, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy (Crown 2016).
OpenDemocracy, ‘NHS Ordered to Share Patient Data with US “Spy-Tech” Firm’ (2021) https://www.opendemocracy.net/en/palantir-peter-thiel-nhs-england-foundry-faster-data-flows/.
Royal College of General Practitioners (RCGP), The Role of the GP in 21st Century Healthcare (2020) https://www.rcgp.org.uk/policy/rcgp-policy-areas/future-role-gp .
The Atlantic, ‘The Philosopher of “Peak Libertarianism”’ (2020) https://www.theatlantic.com/magazine/archive/2020/07/the-philosopher-of-peak-libertarianism/612232/ .
The Guardian, ‘Trump Administration Sued for Giving Medicaid Data to Deportation Officials’ (2 July 2025) https://www.theguardian.com/us-news/2025/jul/02/trump-lawsuit-state-medicaid-data-immigration.
The Guardian, ‘Palantir’s NHS Data Work Raises Concerns over Consent and Commercial Use’ (2023) https://www.theguardian.com/society/2023/jul/15/palantir-nhs-england-data-concerns.
Thiel P, ‘The Education of a Libertarian’ Cato Unbound (13 April 2009) https://www.cato-unbound.org/2009/04/13/peter-thiel/education-libertarian/ .
Wired, ‘Palantir’s God’s-Eye View of the Pandemic’ (2020) https://www.wired.com/story/palantirs-gods-eye-view-of-covid/ .
WHO and OECD, Health at a Glance: Europe 2020 (OECD Publishing 2020) https://www.oecd.org/health/health-at-a-glance-europe-23056088.htm .
Digital Health News, ‘NHS AI Tool Could Deprioritise Certain Patients, Critics Warn’ (2024) https://www.digitalhealth.net/2024/05/nhs-ai-tool-deprioritises-elderly/ .
* Anders Larson is a Lib Dem member who has spent most of his life in the US and EU but is currently based in Norfolk.
Follow @libdemvoice.org on Bluesky
Like us on Facebook
Subscribe to our feed
Sign-up for our daily email digest
The street in Australia named after my great great grandfather
Lib Dems pick Richard Kilpatrick as candidate for the Greater Manchester Mayoral election
Matt Wardman
Peter Martin
expats
15 Comments
At the heart of the proposed reforms is that idea of prioritising the use of scarce resources. In practice this means that where a large number of individuals have exactly the same medical needs that exceeds available capacity, priority will be given to those with the greater chance of surviving any operation, or those most likely to benefit most. This will mean those who are obese, smoke, drink to excess etc will find themselves behind others in terms of priority. Similarly, very elderly patients with a predicted life expectancy of only a few more years anyway, will find themselves behind those who have the same medical need but are are much younger.
I disagree with the article.
In my opinion it is highly desirable to develop AI that can take clinical decisions at least as well as humans. How soon we can achieve this is of course a matter for debate.
With regards to data, the patient data held by the NHS is an incredibly valuable aid to research, and I want to see it exploited as much as possible. I regard opt outs from use of one’s individual data (once it is anonymised) as being highly undesirable, and I would eliminate them. I see the desire to not allow your data to be used for research as extreme selfishness.
My policy would be that if you want to be treated by the NHS, the NHS can use your anonymised data as it wishes.
@Mohammed Amin – firstly, the plan is all-in on the use of unproven AI. It might eventually mature to be an effective medical tool, but it also might not. Where’s the Plan B if it turns out not to be sufficiently reliable and makes more mistakes than human medical professionals, and who is liable for the harm when it gets things wrong? The taxpayer (via the NHS budget) or the (probably US) AI system contractor?
Secondly, I find your views on data and opt-outs rather surprising. Genuine anonymity when it comes to medical data is very difficult to achieve, and there is nothing more personal than that. I think the suggestion of denying health care to those unwilling to share their most private and sensitive data to be quite illiberal, although I suspect we ultimately won’t be given a choice.
> highly desirable to develop AI that can take clinical decisions at least as well as humans
What does “as well as” mean? This is where a lot of assumptions are smuggled in. Error rate? In what? False positive rate? Miss rate? Scope of diagnosis? Some sort of notion of wisdom? Ability to admit lack of accuracy and escalating a case to a human clinician?
What about Accountability or Responsibility? Explaining and justifying a diagnosis? Listing alternative possibilities and reasons why that might be applicable? Discussion of the research base of what you might have and the validation?
I was quite happy lying on the operating table having a chat with humans. I know too much about computers to trust these Artificial Stupidity engines.
And lastly – we do not have Artificial Intelligence, not even close. We have GPT-LLM engines, which is not the same things at all. Mere automation and pattern matching.
So I find article raises some interesting points.
@Mohammed Amin
From the original article:-
“centralises England’s patient data under the control of Palantir, a US surveillance firm with no democratic oversight…”
https://www.bbc.co.uk/news/business-54348456
“US tech firm Palantir, known for supplying controversial data-sifting software to government agencies, has fetched a market value of nearly $22bn (£17bn) in its debut on the New York Stock Exchange.”
“The firm, which launched in 2003 with backing from right-wing libertarian tech investor Peter Thiel and America’s Central Intelligence Agency (CIA), builds programs that integrate massive data sets and spit out connections and patterns in user-friendly formats.”
Who actually runs the software and who has access to the original data -i.e. to our personal medical data?
https://pro.bloomberglaw.com/insights/privacy/state-privacy-legislation-tracker/
“Currently, there are 20 states – including California, Virginia, and Colorado, among others – that have comprehensive data privacy laws in place…….”
“Concurrently, several states have introduced narrow consumer privacy bills that address a range of issues, including protecting biometric identifiers and health data or governing the activities of specific entities like data brokers or internet service providers.”
So there appears to be no comprehensive data privacy law protecting people from misuse of their data.
And would US law protect citizens of other countried anyway?
Not remotely acceptable in my view
As a software engineer working in medical research I’d like to emphasise Nick Baird’s extremely important and often under appreciated point. Medical data can rarely be fully anonymised without destroying its scientific value. For example, MRI brain scans often contain enough facial features to enable identification through facial recognition software. While we can apply “defacing” techniques to brain images, these controls aren’t consistently implemented. This illustrates a broader, systemic problem.
Medical data remains inherently vulnerable to data sleuthing—the practice of exploiting combinations of demographic, clinical, and behavioural information to create unique identifying patterns. Bad actors can circumvent privacy protections through linkage attacks, statistical inference, or cross-referencing with external datasets. Information security and privacy protection cannot be guaranteed once this data is no longer under control of the NHS.
We are gifting an enormous and unique national asset to a foreign company headed by a deeply sinister character. How can this possibly be in the interests of the British people?
Thank you for an outstandingly detailed and pressingly relevant article!
The U. S. A is demonstrably unreliable, internally and externally, and is governed for the financial and status benefits of plutocrats, including Mr. Thief so why give away power and responsibility as proposed?
Might this glossy, shallow policy indicate (further) a lack of Labour forward planning?
Might the L. D. leadership publicise the dangers of further incorporation of our nation into American influence/control which President Trump has clearly stated to be for the purpose of American (elite) benefit?
P.S. Personal research into Mr Thiel by reading about him is informative and, possibly, disturbing.
“In my opinion it is highly desirable to develop AI that can take clinical decisions at least as well as humans. How soon we can achieve this is of course a matter for debate.”
This review of recent research claims ” The use of AI can improve the quality, efficiency, and effectiveness of healthcare services by providing accurate, timely, and personalized information to support decision-making”.
https://pmc.ncbi.nlm.nih.gov/articles/PMC10916499/
I entirely agree that that we should be extremely suspicious of US influence and control.
@Brenda Will
“At the heart of the proposed reforms is that idea of prioritising the use of scarce resources. In practice this means that where a large number of individuals have exactly the same medical needs that exceeds available capacity, priority will be given to those with the greater chance of surviving any operation, or those most likely to benefit most”.
I guess this is one way of prioritising resources. In a world where money is limited, demand for heath spending continue to increase as new treatments and technology becomes available, citizens consider they are overtaxed, and where there are probably greater priorities (net zero, and regrettably defence spending for example) what are the alternatives?
Here is what the BMA has to say, though I expect they will have an eye on the jobs of medics. The Executive Summary is worth a read.
https://www.bma.org.uk/media/njgfbmnn/bma-principles-for-artificial-intelligence-ai-and-its-application-in-healthcare.pdf
Anders’ thought provoking article raises important questions and concerns.Does he or do others have answers to address those concerns and alternative proposals to deal with the reforms to our Health and Social Care system that are rather clearly needed ?
The principle of primary health care through GPs is that the GP can see you as a whole person, which may profoundly influence her/his perception of risk and so on. A target culture and larger, overworked GP practices have largely abandoned this, but it still can be done. Not under this new system. There is also a major issue about citizens who cannot use the app, who are likely to be most in need of care. I do think AI can have a role in diagnosis, but not in decisions on care.
@ Simon Banks “There is also a major issue about citizens who cannot use the app”.
Absolutely right, Simon. There’s something highly illiberal about excluding growing numbers of elderly people from medical, banking and so many other areas of life because they cannot afford or sometimes understand computers, internet, apps and AI systems.
So much for a participatory democracy.
Well said David , l couldn’t agree more.
I really hope that someone senior in the party is going to raise these issues in parliament. I see no signs to suggest that NHS senior management have moved from their usual position of threatening their own staff to maintain silence when individuals or departments are failing badly. Whistleblowers are unlikely to work again anywhere in the NHS.
Most patients or bereaved relatives do not pursue complaints when faced with the management response they get, yet the minority who persist cost the NHS more £3billion a year in fees and compensation. Engaging a remote and unaccountable surveillance business to oversee patient data and influence future access to healthcare is a perfect fit for our broken healthcare management culture. We have already seen how large businesses have ditched customer service with fake chat rooms and unresponsive email addresses.
I would like to see a return to GP practices being managed by their own doctors, rather than the present free for all. Your GP consultation may already be following a tick-list generated by a foreign investor who fancied adding a few dozen UK practices to their portfolio.
Let’s work towards an NHS focused on the patient and empowering their doctors – essentially an inversion of the current priorities.