Palantir and the NHS Data Platform

Wednesday morning, central London. A van pulls up outside Palantir's office. It's been painted to look like a Palestinian ambulance... shot through with bullet holes. Amnesty International and a group called Just Treatment put it there, with a question on the side: why should a company linked to Israel's destruction of Gaza's healthcare system have any place in ours?

That's outside their actual front door?

Their London headquarters. And the same day, September thirtieth, the Guardian reported something that might matter more than the van. Forty-four thousand people have now filed formal legal objections to their health records going anywhere near the NHS platform Palantir runs.

Forty-four thousand.

And all of it is pointed at one date. There's a break clause in this contract. February twenty twenty-seven. Ministers can walk away.

Okay, hold on. Before the drama... what does the software actually DO? Because I hear Palantir and the NHS constantly and I genuinely couldn't tell you.

Right, let's be concrete, because everyone skips this bit. It's called the Federated Data Platform. Picture a hospital trust. The waiting list lives in one system. The operating theatre schedule lives in another. Who's medically ready to go home... that's a third. None of them talk to each other.

Which is why my aunt waits four hours for a bed that was free at ten a.m.

That exact problem. Palantir's Foundry sits on top as a layer. One MP described it in a committee hearing in June as an overlay... you make a query, you don't see the silos, you see a joined-up answer. So the discharge tool tells you, this morning, who's fit to leave and what's blocking each one.

That's... boring. And useful. Boring and useful is usually good.

It's logistics software, deliberately dull. And NHS England's own figures, to the end of May twenty twenty-six, say a hundred and thirty-nine trusts are live, all thirty-five integrated care boards, and the inpatient tool is credited with a hundred and eleven thousand, five hundred and eighty-nine extra theatre procedures.

A hundred and eleven thousand operations is a real number.

Hold that number. Because a campaign group called Foxglove went after it with freedom of information requests, and what came back was... lumpy.

Lumpy how?

Of the reported fall in outpatient waiting lists, eighty-four percent came from one single trust. Chelsea and Westminster. Only sixteen trusts were using that tool at all. And nearly a third of the trusts on the scheduling tool were doing FEWER procedures than before it went live.

Wait. Same software, opposite results? That sounds less like a platform and more like whoever happened to be good at their job.

That's the honest reading. And NHS England has basically conceded it. Their own documentation said it was inappropriate to draw conclusions about cause and effect, because other variables weren't controlled for. In July the statistics watchdog told them to put much stronger caveats on the claims.

So ministers were quoting the headline while the footnote said please don't quote this.

Pretty much. Jim Mackey, chief executive of NHS England, has acknowledged the benefits may not be as big as claimed.

Okay, but what's Palantir's side? Right now they're just the villain in a van story.

And some of their answer is strong. Their UK boss is Louis Mosley. He told MPs the NHS data sits in Amazon's London region and isn't exported anywhere. Each trust gets its own instance, controls who sees it, decides what goes in. Palantir says it only ever touches patient data on the NHS's instructions, and NHS England's line is that suppliers can't access it for their own purposes.

And lock-in? That's the fear I hear most.

Mosley's blunt about it: dependency and lock-in are not the same thing. No technical lock-in, he says, if something better comes along the NHS can migrate. He also points out they won an eighteen-month open tender with thirty independent evaluators.

Says the company that got through the door for one pound.

Which is literally true. The first NHS contract, during covid in twenty twenty, was for one pound. Foxglove calls it land and expand. Then November twenty twenty-three, the big one. Up to three hundred and thirty million pounds over seven years, Palantir leading a group with Accenture, PwC and others.

So the forty-four thousand. Is that just a petition with good PR?

No, and that's why it bites. Thirty-eight Degrees used article twenty-one of UK data protection law, the right to object. Under that, an organisation has to stop processing your data unless it can show compelling legitimate grounds to carry on.

So that's forty-four thousand individual legal demands. Not signatures.

Each with a line saying Palantir's involvement undermines their personal trust in NHS data confidentiality. And separately, the health charity Medact says more than eighty-five thousand patients have complained to their own hospitals.

If I objected, would it hurt my care? Honestly, that's the first thing I'd worry about.

It wouldn't. It doesn't touch your right to treatment. This is mostly hospital operations, waiting lists, theatre lists, discharges. But every opt-out makes the data slightly less complete... which is exactly the cost the pro-platform argument warns about.

So distrust degrades the tool, and then the degraded tool proves the distrust was expensive. That's a trap.

It is. And now back to February. In July the Health and Social Care Committee wrote to the health innovation minister and said drop it, find an alternative. The network of senior NHS data chiefs urged ministers to consider the break clause, saying the platform hasn't demonstrated the value you'd expect after that much investment.

And the government's answer?

Reviewing. Decision expected by December. The health minister Zubir Ahmed told MPs it could be reconsidered if someone else can do the job better, and noted the NHS gets no ownership of the software for all that money. Andy Burnham, prime minister since July, has said nothing in public. Meanwhile Palantir hired Tom Watson, Labour's former deputy leader, this month.

Of course they did.

Here's what I keep circling. Nobody in this fight is arguing about whether joined-up hospital data is a good idea. Everybody wants that. The fight is over who holds the keys... and whether the evidence got oversold to justify handing them over.

And whether any government can say out loud that a three-hundred-million-pound thing didn't work.

Two months to decide. Forty-four thousand letters already on the desk. And one lesson nobody budgeted for... you can buy the software. You cannot procure trust.