Britain and Singapore have made opposite bets on the same problem. One has trained a generative AI on the medical histories of 57 million people and published no clinical outcomes from it. The other has opened a public hospital clinic to slow biological ageing and promised to add three healthy years within a decade.
Both are being sold as preventive care. Only one of them has, so far, shown it can actually change what happens to a patient.
Start with the contradiction that should bother anyone paying attention to national longevity strategy. The NHS Foresight model, developed by researchers at University College London and King's College London with NHS England, was trained on deidentified data from 57 million patients — GP visits, hospital interactions, vaccinations and the national death registry between 2018 and 2023, representing over 10 billion medical events. It is built on Meta's LLaMA 2 architecture and designed to predict more than a thousand future disease diagnoses, including hospital admissions, complications and major events such as myocardial infarction. No other health system of comparable size has attempted a generative AI at this scale.
The technical feat is real, the clinical payoff is unproven, and the ethical scaffolding is provisional.
And yet: the model is still being evaluated. Researchers are testing whether it can retrospectively forecast outcomes in 2023 using data from 2018 to 2022. There are no published metrics. There is no opt-out mechanism. There is no patient redress.
The project operates under pandemic-era research provisions, which permit broader data use than usual — but, as the reporting notes, these permissions are not unlimited. GDPR's application to deidentified-but-not-fully-anonymous health data remains uncertain. This is not a footnote. It is the central tension of population-scale health AI: the technical feat is real, the clinical payoff is unproven, and the ethical scaffolding is provisional.
Now contrast that with what is happening in North East London, where the NHS is running a different kind of experiment — smaller in data scale, larger in demonstrated outcomes.
The AI for Urgent and Emergency Care programme, funded by NHS England as a three-year national demonstrator, covers 2.5 million residents in partnership with Health Navigator and UCLPartners. It uses routinely collected local hospital data to train machine-learning models, adjusted for equity, that identify individuals at rising risk of unplanned admission. A six-month prediction window gives clinical coaches time to intervene. Those identified by the model and screened as eligible by clinicians are offered voluntary enrolment into a proactive coaching intervention focused on activation, motivation and self-management of long-term conditions.
This programme is not speculative. It builds on a decade of evidence, including a five-year, multi-centre randomised controlled trial led by the Nuffield Trust. That trial showed 25% fewer days spent in hospital beds, 34% fewer emergency visits, 24% fewer unplanned admissions, 26% fewer GP referrals to hospital, and no extra workload for GPs.
The interim operational data from the North East London deployment are, if anything, stronger. At roughly 11 months in, over 3,000 patients had been onboarded to clinical coaching. Using propensity score matching, the programme demonstrated statistically significant reductions in unplanned hospital activity — specifically bed days — as well as a mortality impact. The provisional data reportedly exceed the original RCT results, including the 25% reduction in bed days.
More than 50% of those contacted take up the offer of coaching. The testimonials are specific and unsentimental: a single parent who had stopped taking antidepressants because she was too busy with her children, now using a pill reminder app and going to the gym. A patient who didn't know what to say to their GP about an asthma flare-up, now with a same-day appointment alert and a respiratory specialist nurse contact. Patients describe being "found by the NHS."
By March 2027, the programme is on track to support over 9,000 patients. If current results continue, it could save 200 to 300 hospital bed days across North East London. The independent evaluation, led by UCLPartners and delivered by THIS Institute at the University of Cambridge and the health economics consultancy LCP, will assess cost-effectiveness, budget impact and sub-group variation by ethnicity and deprivation.
The North East London programme is not a longevity intervention in the strict sense. It does not measure biological age or epigenetic clocks. But it does something that most longevity medicine has not yet managed at scale: it identifies people at risk, intervenes with a human coach, and shows a reduction in hard endpoints — hospital days, emergency visits, mortality.
Singapore is attempting something more ambitious and, so far, less proven.
The country is investing $350 million in research under the Research, Innovation and Enterprise 2030 framework, launched in May 2026, to extend healthy lifespan by focusing on early detection and prevention of ageing-related decline. The Healthy Longevity Grand Challenge, led by the National Research Foundation in partnership with the Ministry of Health, aims to develop affordable biomarkers and interventions tailored for Asian populations.
The vision is explicit. Within several years, residents could visit a GP to test for early cognitive decline, receive a biological age assessment to find out if they are ageing faster or slower than expected, and receive a precision plan to target physiological decline before illness or frailty sets in. Chong Yap Seng, executive director of the Grand Challenge, compared it to routine vehicle maintenance: "You don't bring your car in only when you have an accident or when it stops working; you advise people to send it for regular servicing. We need that for humans as well."
The clinical infrastructure for this already exists in prototype. In September 2023, the National University Health System launched what it described as the world's first Healthy Longevity Clinic in a public hospital, at Alexandra Hospital. The clinic uses biomarkers of ageing specific to the Singapore population, drawing on research from the NUHS Centre for Healthy Longevity. It offers personalised assessments including arterial stiffness measurements and epigenetic analyses, with interventions spanning lifestyle modifications, medical interventions and targeted therapies. A multidisciplinary team — physicians, nurses, allied health experts and health coaches — crafts individual health plans.
The clinic's stated goal is to amplify the healthspan of Singaporeans by three years within the next decade. Patients can also enrol in ongoing clinical trials of supplements and repurposed drugs.
This is where the Singapore approach diverges sharply from the expensive longevity clinics that have proliferated globally. Brian Kennedy, director of the Centre for Healthy Longevity at the National University Health System, drew the line clearly: "We're not studying $100,000 interventions at the university. We're doing things we can scale, like repurposed drugs and supplements... and biomarker tests that cost $200 or less."
The government's broader strategy reinforces that intent. In October 2024, Health Minister Ong Ye Kung announced $200 million in funding over five years for public healthcare institutions to use AI tools and genomics data for preventive care. By mid-2025, the Ministry of Health will launch a national genetic testing programme for familial hypercholesterolemia. If AI-assisted mammography reading proves effective, it will be progressively rolled out from end-2025. Automated record updating will be implemented throughout the public healthcare system by the end of that year.
Ong was careful to note that well-established regulations serve as "clinical gatekeeping" and ensure that the judgment of healthcare professionals remains in place. "We cannot swing to the other extreme, hemming ourselves in with regulations and rules, and failing to harness the opportunities of AI," he said.
The contrast with the UK's Foresight model is instructive. Singapore is building governance and clinical integration into its AI and genomics rollout from the start. The UK trained a model on 57 million people's records and is still working out whether it predicts anything useful.
That is not a criticism of the researchers. It is a structural difference. The NHS's centralised data architecture made Foresight technically possible. But centralised data does not automatically produce accountable clinical pathways. North East London works because the AI is paired with human coaches, voluntary enrolment and an independent evaluation measuring hard endpoints.
Singapore's Healthy Longevity Clinic works, so far, because it is embedded in a public hospital with a research unit feeding it validated diagnostics. The $350 million grand challenge is designed to produce the biomarkers and interventions that the clinic will eventually deploy at population scale through Healthier SG, the national preventive care initiative.
The unanswered question for both countries is the same. Prediction is not prevention. Identifying risk is not the same as changing it. The North East London programme has shown that AI plus human coaching can reduce hospital days and mortality in a defined population. The Foresight model has not yet shown that identifying a thousand future diagnoses changes a single one of them.
Singapore's bet is that biological age assessment and precision geromedicine will prove similarly actionable. That bet is not yet settled. The clinic opened in 2023. The grand challenge launched in 2026. The three-year healthspan goal extends to the mid-2030s.
What both programmes share is a rejection of the status quo. The NHS is shifting from reactive to anticipatory care, from hospital to community. Singapore is shifting from sick care to preventive care. Both are trying to shorten the years spent in poor health.
The difference is that one has published interim data showing it works. The other has published a plan.