Ditch Palantir

NHS England awarded the Federated Data Platform to a consortium led by Palantir — with Accenture, PwC, NECS and Carnall Farrar — in November 2023. The deal is worth up to £330m over a maximum seven-year term. The health minister told the Commons that around £210m had already been spent by spring 2026, before the contract has even reached its first break point.

Ministers sell the platform as a modernisation, but the people who use it are unconvinced. An internal NHS briefing seen by reporters called it "slow and clunky" with a "poor user experience", noting that a task taking 30 seconds in the old system took four to five minutes in Palantir's. A Freedom of Information investigation found harder evidence still: of the trusts using the platform's inpatient-scheduling tool, about a third were carrying out fewer operations than before they adopted it — some 9,000 fewer in total. Palantir disputes the comparisons, pointing to extra activity and reduced discharge delays elsewhere.

Ministers insist the NHS keeps what it pays for. The contract, and Palantir's own standard terms, tell a different story — and even the work already done inside the platform may not survive the end of the deal.

Price discipline comes from competition, and here there was almost none. The platform grew out of the COVID-19 Data Store, handed to Palantir in 2020 for £1 under emergency rules with no open tender, then renewed for £23m — again without evidence of competition — and disclosed only under legal pressure. The £330m deal that followed was published so heavily redacted, with roughly 200 pages blacked out, that its terms cannot be scrutinised either.

A tool that clearly worked would sell itself. Instead, uptake of the platform has been pushed with money, pressure, and — some staff say — threats.

The NHS is not one organisation but many: trusts run their own systems, close to the clinicians and patients who use them. Forcing a single central platform on top both creates a single point of failure and cuts across the local data systems staff already trust. A Commons debate in April 2026 heard that infrastructure this critical should be fully owned and controlled by the NHS — the opposite of what this contract delivers.

The clearest sign that the safeguards are weak is that they have already been exceeded. Within the platform's first years, access to patient data widened in ways the public was never told about — and in at least one case the regulator only found out after the fact.

NHS England's assurances are real: it says Palantir is a data processor, not a controller, cannot commercialise NHS data or train its own models on it, and that all processing happens in the UK. But a contract clause is only as durable as the legal system standing behind it — and the system standing behind Palantir is American.

Defenders of the deal offer three reassurances. Each is true as far as it goes — and each is answered by something already on the record.

The concern is not confined to privacy campaigners. Two Ministry of Defence officials reportedly told journalists that Palantir's deepening knowledge of the British state is itself "a national security threat to the UK." That reframes the whole issue: a supplier that understands your systems this intimately is a strategic vulnerability, not merely a data-protection question.

There is a live precedent for exactly this risk. After President Trump sanctioned ICC chief prosecutor Karim Khan by executive order in February 2025, Khan lost access to his Microsoft email and moved to a Swiss provider; reporting indicated the court was warned that either his access would end or the whole organisation's email would be cut off. Commentators treated it as a demonstration of how dependence on US technology can be switched into leverage for political ends.

Even as ministers say they want a more sovereign technology base, Palantir's UK public-sector footprint has kept expanding through 2026 — a Financial Conduct Authority trial, a quarter-billion-pound follow-on Ministry of Defence agreement, and a contract to run police firearms licensing. Each new deal hands the company a deeper map of the British state, and makes the dependence harder to unwind.

Palantir's chairman and largest early backer has spent more than a decade stating, funding and promoting the view that democracy is a problem to be managed rather than a value to be defended.

This is not only the chairman. The chief executive has put a militarised, civilisation-ranking worldview at the centre of how the company describes itself in public.

It is worth being precise. Palantir was not founded as a vehicle for strongman politics: it began in 2003 as a post-9/11 data-analytics firm, backed early by In-Q-Tel, the CIA's venture arm, and Thiel is understood no longer to run it day to day. The argument is not about origins but about values and trust — the people who own and lead it hold openly anti-democratic views and now advertise a hard-power ideology as their mission. The Health Secretary has himself acknowledged that the founders' politics are a legitimate concern.

The platform runs on a 3+2+1+1 structure: a three-year initial term ending March 2027, then extensions the Department of Health and Social Care must actively choose to trigger. Do nothing, and the contract simply ends. Ministers have said the decision will be taken this year, and the minister told the Commons the break clause is the moment to "evaluate it and find out if there are other providers that can do the job better." This is a real decision, and one that MPs can influence — now, not after it is made.

While attention is on the NHS deal, Palantir's reach across government is widening on other fronts, often awarded directly rather than competed. Every new contract is another reason walking away will be harder next year than it is this year:

• Ministry of Defence — a £240.6m follow-on enterprise agreement, signed in December 2025 as a direct award with no competition.
• Financial Conduct Authority — a £375,000 AI "enterprise search" trial across the regulator's entire data lake, run in early 2026, with a wider rollout still open.
• Police, all 43 forces in England and Wales — a £9m contract to run the national firearms licensing system, awarded in May 2026.
• Metropolitan Police — a £50m analytics deal blocked by the Mayor's office for lack of competition, which Palantir is now challenging in the High Court.

Waiting is not neutral. Every month the deals continue, more clinical and operational work moves inside a platform it cannot easily be exported from, and every new department contract deepens the dependence. The moment of maximum leverage is now, before the extension is triggered; afterwards, the government negotiates from inside the lock-in.

The documented pattern is consistent: embed analysts inside the organisation, become the system of record, widen data access over time. At contract award, specialists warned the real danger was military-style "mission creep" — the argument that the job could be done a little better with a little more data. Once that has run its course, what stands between the data and its misuse is not a technical control but our trust in the company holding it — the same company whose owners reject democracy and whose stated mission is hard power.

Sovereignty over the data would no longer be ours alone. As the earlier sections set out, US law can compel an American company to hand over data wherever it is held; the current administration has already turned tech dependence into leverage against the ICC; and even Microsoft's own lawyers say they cannot guarantee otherwise. To entrench Palantir is to bet that this capability will never be turned on the UK.

The exit is the weakest point of all. Palantir's own terms convey "no ownership rights" to the customer; the NHS contract leaves it with no software and no intellectual property; and switching off tools already woven into live services — cancer pathways among them — would break those services. In June 2026 a Commons committee called the UK's growing reliance on Palantir "an unacceptable point of weakness" and warned the country was "seriously exposed" to vendor lock-in. Dependence, once complete, is hard to tell apart from control.