Palantir NHS objections pass 44,000 as AI coding inflates US hospital bills

More than 44,000 people have now filed formal objections under UK data protection law demanding NHS England stop the Federated Data Platform (FDP) from handling their personal health data, according to The Guardian, which reported the figure on 1 October. The platform runs on technology from US company Palantir.

The objections invoke article 21 of the UK General Data Protection Regulation and say the involvement of Palantir "undermines my personal trust in NHS data confidentiality". Under those rights, the NHS must stop using the data unless it can show a compelling reason to continue.

That number matters because it is not a petition. Petitions have already gathered hundreds of thousands of signatures, the same Guardian report says, but a formal objection under data protection law puts an obligation on the data controller. Andy Burnham will be required to decide whether to back or axe Palantir within months, and two House of Commons select committees have called on Labour to use a break clause in the seven-year deal, which comes into force in February 2027. The contract is worth £330m.

What the platform is supposed to deliver

Palantir says the FDP connects fragmented health service data, from surgery schedules to discharge planning, and that it is helping cut waiting lists and free beds. Its UK and Europe executive vice president, Louis Mosley, has accused critics of suffering from "Palantir derangement syndrome". A company spokesperson said trusts using the software recorded 117,000 additional operations, a 14.3% reduction in discharge delays for long-stay patients and a 5.6% improvement in the number of people finding out whether they have cancer within 28 days.

Campaigners point to Palantir's work for the Israeli military and for Donald Trump's immigration enforcement. Mosley has said its technology was used for locating hostages, not military targeting, in Gaza. When challenged last year by a protester alleging "Palantir kills Palestinians", chief executive Alex Karp replied: "Mostly terrorists, that's true." On Wednesday, Amnesty International and the patient campaign group Just Treatment parked a van styled like a shot-up Palestinian ambulance outside Palantir's London HQ. Palantir says patient data can only be handled on NHS instructions and remains in the UK.

NHS England chief executive Jim Mackey has acknowledged concerns that the benefits to public health may not be as significant as claimed. That admission sits awkwardly next to the promotional numbers, and it is the kind of gap that procurement reviews tend to find.

The billing side: AI as a cost multiplier

Across the Atlantic, the argument has moved from trust to money. The Blue Cross Blue Shield Association (BCBSA) estimated that hospitals' use of AI-assisted medical coding contributed to close to $1 billion, or $942 million, in additional costs for its health plans between 2023 and 2025, CNBC reported on 1 October. BCBSA said much of the increase came from secondary diagnoses that moved patients into higher-paying reimbursement categories, and described those diagnoses as types that "may be derived from single laboratory values, making it particularly well suited for detection by AI tools".

Luke Chalker, BCBSA's senior vice president of product and data science, told CNBC that roughly 70%, or $653 million, of the billing identified was tied to additional diagnoses that came with no change in care. He stopped short of attributing the entire increase to AI, saying multiple factors contribute to coding intensity but that the findings suggest AI-enabled coding and documentation tools are playing a role. BCBSA wrote that the growth in "complex coding" arrived during a period when 60% of hospital systems began using AI coding tools, and concluded there is "a clear disconnect between coding and treatment".

Christopher Whaley, a health economist at Brown University who studies hospital coding, told CNBC that AI appears to be "accelerating, and in some sense making it easier to capture, the existing and underlying billing incentives that are in the system". He added that in many cases the additional diagnoses are legitimate and simply were not captured before, but that some conditions "clinically just don't really matter and don't influence the patient's care" while still allowing another billing code to be applied.

Investors keep funding the admin layer

The money keeps moving regardless. Endpoints News reported exclusively on 1 October that Parakeet Health raised $10 million to automate patient outreach by phone, text and email. The same outlet reported that Medicare Advantage insurer and medical group Devoted Health announced nearly $1.2 billion in fresh funding after enrolling hundreds of thousands of new senior members.

In Europe, Portuguese insurtech FRANK raised €2.9 million in a round led by Armilar and Start Ventures, Tech.eu reported on 1 October, for an AI assistant that logs into insurer portals on behalf of brokers. The company says it is used by more than 350 insurance agencies and brokerages, over 1,500 users in Portugal, and handles 70,000 back-office tasks a month. Tech.eu also reported that Berlin-based voice AI company Deepslate raised €7.7 million in a seed round led by 42CAP, claiming a 440 millisecond response time in an independent Artificial Analysis benchmark as of September 2026.

The pattern is consistent: capital flows into tools that sit between clinical work and payment. What is thinner is evidence that these tools reduce total cost. Endpoints News framed the question bluntly on 1 October, noting that for all the talk about AI cutting healthcare costs, early signs point the other way.

Spending plans and the research base

Omdia's outlook, published by Light Reading on 30 September, predicts that 59% of organisations expect AI budgets to rise by 10% or more in 2027, and that enterprises will judge AI investments by ROI and productivity gains rather than technical superiority. Already, hardware delays are affecting 60% of PC channel partners, the analyst firm said. More than 100 countries are pursuing digital sovereignty initiatives, a figure that maps directly onto the NHS debate about where data sits and who operates the platform.

The evidence base for clinical AI is still catching up. On 28 September, researchers including Silas Ruhrberg Estévez, Russ B. Altman and Mihaela van der Schaar posted a paper to arXiv describing EHRFlow, a flow-matching framework that conditions on encoded patient history and, on datasets covering more than one million patients, improved horizon-averaged top-5 clinical-code accuracy over autoregressive and history-independent baselines. It is a code forecasting benchmark, not a billing study, but it is the sort of method that determines what future coding tools can detect.

Health systems are also being pushed to look beyond reproduction in how they study disease. Science devoted a special issue to women's health published on Thursday, edited by an all-female team, which flagged that women remain poorly represented in clinical trials and that women experience functional pain at about twice the rate of men. The editors named funding cuts affecting women-specific research as an urgent motive behind the issue.

And the data collection itself is shrinking. STAT reported on 1 October that the CDC's National Center for Health Statistics proposed an overhaul of the National Health Interview Survey in August, cutting the adult questionnaire from 482 questions to 150 and removing questions about hearing aids, fatigue, cognition and mobility supports such as wheelchairs and scooters. Roughly 25,000 households will receive a mailed questionnaire. A recent paper co-authored by CDC statisticians suggests past surveys may have missed 75% of people with intellectual and developmental disabilities.

What to watch

Two clocks are now running in parallel. The first is contractual: the FDP break clause becomes available in February 2027, and 44,000 legal objections give NHS England a documented set of people who have asked it to stop. The second is financial: BCBSA's $942 million estimate covers 2023 to 2025 and will be tested against whatever insurers find in 2026 claims.

Neither side has published a clean counterfactual. Palantir points to operations and discharge metrics. BCBSA points to coding intensity. Whaley's framing, that AI accelerates incentives already present, may be the most useful thing said this week, because it suggests the technology is not the cause of the cost problem, only a faster way of expressing it.