A British health minister has told a parliamentary committee that public distrust of Palantir is now measurable in the NHS's own data. In a letter to Layla Moran, chair of the Commons health committee, Frith wrote that distrust of the American defence and healthcare technology company could affect how willing people are to hand over their data — and attached a number. Between mid-May and mid-July, a two-month window, an additional 60,000 patients used the national data opt-out to withdraw permission for their personal medical information to be used. Frith linked the increase to heavier media coverage of Palantir's role. It is the first time the government has priced the political cost of its supplier in units of lost data.
The same letter quietly conceded something bigger. In July, the health minister told parliament that NHS trusts would be ordered to use certain components of the federated data platform, which runs on Palantir technology. Frith now says there is no mandatory requirement to use the FDP and that organisations may choose an alternative.
Read those two paragraphs together and the shape of the retreat is clear. A £330m, seven-year contract for a national platform was justified on the logic of a national platform: one system, everyone on it, network effects across trusts. A platform each trust may decline is a different product with a different value case, argued hospital by hospital. Nothing about the contract changed this summer. The department's confidence that it could compel people onto it did.
Doctors and patient organisations are pressing Labour to use the break clause in the seven-year agreement. They point to Palantir's work with the Israeli military and with the immigration agency ICE under Donald Trump, and they dispute the company's claims that the contract is good value for the NHS.
Those claims are specific. Palantir and its supporters say the AI-based technology has helped the NHS cut surgical waiting lists. A company spokesperson said that trusts using the platform carried out 110,000 more operations, reduced discharge delays for patients on long hospital stays by 15%, and raised by 6.8% the number of people who learned within 28 days whether they had cancer. NHS England has reported comparable figures but stressed that they do not establish cause and effect. The UK Office for Statistics Regulation is investigating where the numbers came from and how they have been interpreted.
That is an awkward position for the government. It is defending a contract partly on efficacy figures that its own statistics regulator is currently examining, while the body that actually holds the data declines to claim the results were caused by the software.
Palantir's public sector work in Britain has generated friction before. Ministers had to apologise after it emerged that company staff had been given access to patient data from which individuals could be identified. Sadiq Khan, the mayor of London, blocked a £50m contract between the Metropolitan Police and Palantir for AI support in investigations, saying Londoners want public money to go only to companies that share the city's values. Palantir is suing over the mayor's decision.
Frith called the rise in opt-outs moderate, and said it would be monitored. Opting out does not stop a person's medical history being used for their direct care, including operations in real time; it removes the data from research and planning. The minister warned that the benefits of the ten-year health plan may prove impossible to realise if patients stop sharing their data.
Foxglove, the tech justice group, welcomed the minister's acknowledgement of public distrust. Tom Hegarty, its head of communications, said doctors and patients had warned about the threat to the NHS and that tens of thousands of people had opted out of data use in recent months. He added that the concern is warranted given that Palantir founder and Trump ally Peter Thiel has publicly argued that the NHS makes people sick.
My own view is that "moderate" is doing more work in that letter than it can bear, and not because the number is large. It is that the letter gives no baseline. Sixty thousand opt-outs in two months is a rate with no denominator attached and no prior period to compare it against, which makes the minister's characterisation unfalsifiable from the outside. A government genuinely confident the trend is mild would publish the series. And the causal claim runs in a direction that should worry ministers more than it seems to: Frith attributes the rise to media attention rather than to anything the department or the supplier did, which is a theory in which the remedy is quieter coverage.
The letter also says nothing about what happens next to those 60,000 records, and I suspect that is the asymmetry that matters. An opt-out is an action a person takes once, on their own initiative, about something they read; nothing in the system prompts them to revisit it when the coverage stops. Palantir's contract has a break clause and an end date. The withdrawn permissions have neither. If the government changes supplier tomorrow, it inherits a research base with a hole in it that no procurement decision repairs.
Moran, the Liberal Democrat MP, called the softening on mandatory FDP adoption a deserved change, and said the government should switch supplier and not extend the current contract — adding that, given how long the process takes, ministers need to decide quickly rather than leave it until February. That deadline is the real test of the letter. A department that has already dropped the mandate has conceded that the platform must win trusts on merit. The next question is whether it is willing to find out what happens when the same standard is applied to the contract itself.