Peter Geoghegan,Lucas Amin: ‘This looks absolutely rubbish’

Peter Geoghegan and Lucas Amin on Palantir and the NHS

There’s no sign over the door of Palantir’s UK headquarters, an eight-storey neoclassical building in Soho Square. The windows are lined with thick reflective glass. When we tried to take some photographs outside recently, a man with a lanyard told us to stop. He wouldn’t be drawn on his employer: ‘It’s just a tech company.’ Two independent sources who have attended meetings there with senior Palantir staff in the last two years told us they had to sign non-disclosure agreements just to enter the building.

Such wariness may be understandable. In 2023 two hundred people blockaded the offices in protest against the decision to award Palantir a £330 million contract to supply NHS England with data integration and analysis software. Last October activists from a group called Shut the System doused the façade of the building in red paint – a reference to Palantir’s partnership with the Israeli Ministry of Defence for ‘war-related missions’. Human rights organisations allege that Palantir’s software is integrated into the IDF’s targeting systems in Gaza. (The company has described the claims as ‘misinformation’.)

Palantir hasn’t altered its behaviour in the face of such criticism. Peter Thiel, the firm’s Trump-supporting co-founder, talks enthusiastically of melding technological progress with nationalist politics. Alex Karp, Thiel’s business partner and classmate at Stanford, who has a PhD in social theory from the Goethe University Frankfurt, has joked about using a drone to spray sceptical market analysts with ‘light fentanyl-based urine’. In April, a 22-point manifesto by Karp heralding the dawn of AI-enabled US military supremacy was posted on Palantir’s X account. As the company’s founders embrace Trump’s increasingly authoritarian second term, its market capitalisation has risen to $320 billion, 150 times earnings. The Economist has speculated that ‘Palantir might be the most overvalued firm of all time.’

Palantir’s expansion in the UK has been rapid. A decade ago the firm was largely unknown here. It now holds public contracts worth hundreds of millions of pounds in everything from financial regulation to local government. Its head of UK operations, Louis Mosley (grandson of Oswald), appears regularly in the British media. Although usually more measured than his US bosses, at last year’s Alliance for Responsible Citizenship conference he warned that Western civilisation was in an existential battle against a ‘distributed idea suppression complex’ consisting of ‘armies of fact-checkers and experts’, lawyers, activists, academics and journalists. Organised opposition to Palantir has increased in Britain, but Labour ministers continue to stress the importance of its technology to the running of a modern state.

In recent months the investigative newsletter Democracy for Sale, where we both work, has been looking into Palantir’s business in the UK, and particularly its contracts with the NHS. We have obtained scores of internal documents and spoken to sources who have worked with Palantir, its suppliers and its software. Behind the company’s bombastic self-image, we found there was often a much more prosaic reality: problems with software, pressure from management, staff resistance, but also an unrivalled level of political access.

Palantir began in 2003, in a small office above a sushi restaurant in Palo Alto. The name comes from one of the indestructible balls made by elves in The Lord of the Rings: in Tolkien’s Elvish it means ‘one that sees from afar’. The company was founded on a simple premise: governments and public agencies hold a lot of information, but it is siloed and poorly organised. If properly collated and arranged, the data could be employed to enormously powerful effect. For Thiel, Palantir has always been an explicitly political project. In the wake of 9/11 he was growing leery of globalisation, immigration, democracy itself. As he became increasingly preoccupied by the threat of Islamic terrorism, he pitched Palantir as a security product: its cutting-edge software would mine government data and use network analysis to identify terrorists. Ethical qualms about human rights and data privacy were a secondary consideration at best. Palantir’s first outside investor was the CIA’s private venture capital arm, In-Q-Tel. The agency was its first customer. Thiel once joked that ‘the CIA is a front for Palantir.’

Early iterations of Palantir software took in data from multiple sources and allowed intelligence analysts to search, link and visualise connections across people, places, events, and even phone records and watchlists. After Palantir launched its first product, Palantir Gotham, in 2008, ranks of lobbyists and paid boosters helped push it to government and intelligence analysts. From early on, Palantir had some of the biggest lobbying firms in Washington working on its behalf, as well as well-connected political insiders: Reagan’s national security adviser John Poindexter, the Clinton strategist James Carville and, later, the disgraced general Mike Flynn, who was Trump’s first national security adviser. Former US government and intelligence officials were hired to fill executive and advisory roles. In 2009, the company won its first big commercial contract, with JPMorgan Chase.

But Palantir’s technology rarely justified the hype. Palantir talked up its contribution to the CIA’s tracking down of Osama bin Laden – it claimed that its software was used to sift and integrate data from disparate intelligence streams (bank records, travel logs and so on) into a single platform, helping analysts to identify the compound in Abbottabad – but few in national security or intelligence circles bought this. A veteran intelligence analyst quoted in Max Chafkin’s biography of Thiel, The Contrarian, dismissed Palantir’s intelligence prowess as ‘this totally fake thing’.* As the largesse of the war on terror gave way to Barack Obama’s ‘light footprint’, Palantir’s early promise seemed to be evaporating. It was losing tens of millions of dollars. Corporate clients began falling away. By 2014, Karp had scaled down its US commercial business and relocated to Europe.

London was an obvious destination for a US tech company with designs on hefty public contracts. The market was large if crowded. Palantir had some connections – it had provided software to GCHQ – and was soon spending heavily to build its contacts book. In 2015 it won its first published public contract, a £734,834 deal to provide the Government Digital Service with an analytics platform. Palantir was willing to spend more on securing political access than almost all of its British and Continental competitors. It made donations to London think tanks and hosted parties at Davos. Senior civil servants and military men were given lucrative advisory roles. As it had in Washington, Palantir built ties with politicians. Mosley, hired in 2016, was a Tory councillor in Kensington and Chelsea who had previously worked for Rory Stewart. In 2018 Peter Mandelson’s lobbying firm, Global Counsel, was brought in, the New York Times wrote, to position Palantir ‘as a respectable partner to the British government, not a shadowy American military contractor’.

Global Counsel was founded in 2010 by Mandelson and Benjamin Wegg-Prosser, a former adviser to Blair. Its early clients included the now sanctioned oligarch Oleg Deripaska, Uber and Vodafone. But Mandelson was unhappy that the company wasn’t making more money. In emails with Jeffrey Epstein – correspondence that destroyed Global Counsel when it was released earlier this year – Mandelson complained that he needed bigger clients. Palantir fit the bill. Its monthly retainer was worth more than £30,000, making it one of Global Counsel’s most lucrative accounts, according to two sources familiar with the arrangement. Mandelson took a personal interest in the firm, attending dinners Global Counsel arranged for Palantir and briefing senior executives. ‘There were some clients Peter never saw. Palantir definitely wasn’t like that,’ a former Global Counsel employee recalled. ‘Peter had the contacts across the political spectrum. His Rolodex was a big part of the sell [to Palantir],’ another said.

Global Counsel marketed itself as an advisory firm, not a lobby group. Although it did produce political intelligence briefings for Palantir and, later, rebuttals against criticism of the firm, former Global Counsel staff told us that one of the primary goals of the contract was to get the company in front of British politicians and policymakers. ‘UK politicians are a lot more easily impressed than American politicians. It costs a lot less to influence them,’ a former Global Counsel employee said. Another recalled that Global Counsel ‘would often do these dinners, like “how to maximise data innovation in the NHS”. And they wouldn’t say “sponsored by Palantir”, but they would be there, and then that would allow Palantir to talk about what they’re doing.’ Under the UK’s weak lobbying laws, policy briefings are not automatically classified as paid advocacy, which meant that Global Counsel didn’t have to declare Palantir as a client on public registers. ‘Global Counsel made sure that everyone knew not to contact ministers and senior civil servants directly,’ since this would have to be registered, a former employee told us. ‘The register is so narrow, and with such a narrow definition of what constitutes lobbying, you really have to fuck up to have to register somebody.’ Global Counsel didn’t declare the company as a client until 2022, and only did so then after a junior employee mistakenly made representations on Palantir’s behalf directly to a senior government official, triggering the disclosure.

The Palantir account was also handled with a high level of secrecy within Global Counsel. It was always referred to by a code name, Project Onion, even on internal systems. Palantir wasn’t a universally popular client among those who knew about the contract. A former employee said they tried to avoid the account after doing some work for Mosley. ‘I found him very difficult,’ the source said. ‘Demanding, aggressive, ruthless, power-hungry.’ (Someone who has known Mosley for more than a decade gave us a different view: ‘He is definitely a very ideologically motivated person. He’s also very personable. He’s very charming. He’s very clever.’)

By June 2018 Palantir had won its first seven-figure UK contract: a pilot project for a naval personnel management system for the Ministry of Defence. But Palantir, and Global Counsel, had a much bigger target. ‘The NHS was the holy grail,’ according to a former Global Counsel employee. At this point, the UK’s health data was held in thousands of places – doctor’s surgeries, NHS trusts, pharmacies, third-party providers – and in a variety of file formats. Creating a single digital system for NHS data had been a government objective for decades, but previous efforts had ended badly. Blair’s National Programme for Information Technology, launched in 2002, aimed to create a single electronic record for every patient. It was shut down nine years later, ‘having mostly failed to achieve its goals’, according to an independent report commissioned by the government, and hugely overrun its budget, at a total cost of £12.4 billion. In 2013 the Public Accounts Committee called the programme ‘one of the worst and most expensive contracting fiascos in the history of the public sector’. ‘We wasted a lot of money on big shiny American solutions that didn’t really work,’ a senior IT director in the NPfIT, Joe McDonald, who was fired after raising concerns about overspending on tech contractors, told us. A second attempt to centralise and share patient records electronically, known as care.data, collapsed in 2014 after it was revealed that the Department of Health and Social Care had sold the records of 47 million patients to an insurer.

Other initiatives have been more successful. NHS Spine, which was built using low-cost open-source tools and has been in use since 2014, enables 26,000 organisations including hospitals, GP practices and pharmacies to access patient data quickly and safely and to exchange information. But in the age of big data and AI, there remained much to be done. Joining up information and making it easily accessible to the right people could save resources and facilitate new research.

Despite having no real history of working with health data, Palantir began positioning itself as the go-to expert and Global Counsel started hiring Westminster insiders who had contacts in healthcare. Nicola Blackwood, a former Conservative health minister focused on NHS innovation, joined as a part-time adviser in 2018. She left the firm the following year when she returned to the Ministry of Health as a Tory peer, but after becoming chair of Genomics England in 2020 received payments from Global Counsel to attend events where the guests included Palantir.

During 2019 Palantir’s lobbying was persistent, but the company was careful to avoid triggering disclosure rules. That July, the evening before the launch of the NHS’s new technology arm, NHSX, Mosley hosted a meal attended by David Prior, chair of NHS England. Documents obtained by the Bureau of Investigative Journalism show that Mosley emailed Prior the next day, thanking him for coming. Prior replied that evening: ‘Louis, Thank you for hosting such an interesting dinner and also for the water melon cocktails! If you can see ways where you could help us structure and curate our data so that it helps us deliver better care and provides a more insightful data base for medical research do be in touch.’ A couple of weeks later, the former head of MI6, John Sawers – one of dozens of former politicians and senior civil servants hired as advisers by Palantir – facilitated a meeting between Karp and John Manzoni, permanent secretary at the Cabinet Office. In subsequent correspondence with the department, Sawers hailed the benefits of Palantir’s ‘unique software’. A year later the Cabinet Office gave Palantir a £27 million contract.

The Covid pandemic accelerated Palantir’s advance in the UK. In August 2019, Boris Johnson and Dominic Cummings had an off-the-books meeting with Thiel in Downing Street. Six months later, Palantir was one of the thirty-odd tech companies invited to Downing Street and asked to aid the national pandemic response. Palantir, with support from Google, Microsoft and Faculty (the AI firm used by Cummings during the Vote Leave campaign), was charged with building a repository of datasets to manage and co-ordinate NHS England’s effort. This ‘NHS datastore’ would run on Foundry, a software system built by Palantir to integrate data from disparate sources. Palantir charged the government just £1 for the initial four-month contract. Within six months the fee had risen to £1.5 million for six months’ work. In sales this is known as ‘land and expand’: get a foot in the door, then start upselling. By December 2020 Palantir had signed a £23.5 million contract with the Department of Health.

The company got more work in other sectors too. Between April 2020 and March 2022 the Ministry of Defence awarded Palantir three data integration contracts worth £17.6 million. That November, Palantir hired Polly Scully, a former strategy director at the MoD, as its UK defence lead. The following month it was given a three-year contract worth £75.2 million to work on British military operations and intelligence. It also signed deals with various police forces and businesses such as BP, PwC and Airbus. As with Covid, unexpected shocks provided new opportunities. In March 2022, shortly after Russia’s full-scale invasion of Ukraine, Palantir struck a deal with what was then the Department for Levelling Up, Housing and Communities to provide the IT system for a website that would help find homes for refugees from the conflict. The initial cost was zero. But in another example of ‘land and expand’, the contract was renewed in September 2022 for £4.5 million, then extended a year later for another £5.5 million. (The National Audit Office later found that the government had paid over the odds for this software and that councils found it ‘confusing to use’. The department ‘acknowledged it should plan to move away from Palantir given DLUHC perceived that the costs could be higher than other solutions’. Palantir told us that it was ‘proud to have stepped up to support the Homes for Ukraine scheme and, in particular, to have stood up a solution in just nine days, which enabled the safe resettlement of more than 120,000 refugees’.)

A major NHS contract remained the company’s chief ambition and Palantir continued its campaign to strengthen its influence in the health service and in Whitehall. A key goal was to neutralise the growing criticism of its role in the British state, and in healthcare in particular. It strengthened its political clout by hiring more well-connected insiders. Alongside Global Counsel, Palantir worked with Fleetwood Strategy, a lobbying firm headed by Isaac Levido, an Australian political strategist who had masterminded Johnson’s successful 2019 general election campaign. Levido was later hired, less successfully, by Rishi Sunak to work on general election strategy. (When Sunak visited the US in June 2023 he went to a baseball game with Karp.)

Palantir also poached top executives from the NHS. In June 2021, Paul Howells, a former ‘transformation lead’ at NHS England, became Palantir’s health and care director. The following April, Indra Joshi, who had overseen the development of the Covid datastore for NHSX, joined the company; she was joined by Harjeet Dhaliwal, a former deputy director of data services at NHS England. Around the same time, NHS England announced its intention to build what it called a ‘federated data platform’. The FDP was in some ways a reprise of previous efforts to reform data-sharing from the top down. It had three broad purposes: to join up health records from different platforms across the NHS, to use this data to build specialist apps that could tackle clinical problems – such as cutting waiting lists for surgery or speeding up cancer treatment – and to build a powerful new analytical platform that would enable researchers to spot patterns and opportunities for improvement.

Palantir was the frontrunner for the FDP contract from the off. Few healthcare experts disagreed that data-sharing across the NHS needed to be improved, but many doctors had misgivings about the prospect of some of Britain’s most sensitive data being handed over to the company. Its year of lobbying had already tilted the tendering process in its favour, however. NHS officials spoke of their wish to create a ‘single source of truth’ – one of Palantir’s own slogans. The Financial Times has reported that the invitation to tender even included screenshots of Palantir’s systems – with the logo obscured – and asked for features similar to those the firm already offered. ‘It’s all about setting the terms of the tender,’ a former Global Counsel employee explained. ‘By the time the contract is published the deal is done. Nobody else was going to win it but Palantir. It was all designed to be won by Palantir.’ Shane Tickell, a health technology strategist who led a consortium of British businesses in a rival bid, told us that Palantir ‘had all these people to open doors and lobby and pitch long before the procurement started. And it’s sharp practice. It was good practice. It was great selling. Not saying they did anything illegal, but they used every form of influence.’

In November 2023, NHS England awarded the FDP contract – worth up to £330 million over seven years – to a consortium led by Palantir, alongside the management consultancies Accenture, PwC, Carnall Farrar and NECS. It was the first NHS contract Palantir had won after a competitive tendering process. NHS England told us that it ‘conducted a rigorous, competitive, fair and transparent procurement process over eleven months for a supplier of the NHS Federated Data Platform, in line with public contract regulations, where any eligible supplier could respond’. A few months later, in February 2024, NHS England’s chief data officer, Ming Tang, was the guest of honour at a Global Counsel dinner. A representative from Palantir was also present.

On a Friday afternoon in May 2024, NHS England leaders held a meeting with two hundred data analysts and engineers from across the health service. Unusually, the meeting wasn’t held at NHS England’s headquarters or a conference centre, but at Palantir’s offices in Soho. The rollout of the FDP had officially begun less than two months earlier, and already there were concerns about the platform’s effectiveness. From a stage in Palantir’s basement auditorium, Tang and Vin Diwakar, NHS England’s clinical transformation director, addressed their staff. ‘We all got sort of slapped down and told: you better get on with this. Or you know, things will happen,’ one person present recalled. After the lecture, Palantir officials opened up some ‘tech booths’ to give demonstrations. Not everyone was impressed. ‘I was sitting there thinking, it just doesn’t look very good,’ a source said. At one point a senior data analyst stood up and said: ‘We’ve done all this already. This looks absolutely rubbish.’

The FDP looks much as it did then, with the NHS blue and white colour scheme. Once NHS staff log in, they can access dashboards showing health data from around the country and bar charts that list daily statistics: hospital room bookings, vaccination rates, admissions. They can see forecasts for A&E attendance or cancer appointments, and use a suite of almost a dozen apps built by Palantir, which draw on this data to help clinicians and administrators tackle problems such as identifying extra capacity in operating theatres, finding spare beds in hospital wards and tracking patients using ‘real-time hospital journey monitoring’ to improve outcomes.

As of June this year, 139 trusts – 64 per cent of the total – were ‘live’ on the FDP platform, according to NHS England figures. Appearing before a parliamentary select committee last month, the Department of Health’s digital, data and technology officer said: ‘I look at the number of 139 acute trusts having signed up to use it as an indicator of whether they want or need it.’ But a more reliable way of measuring the success of implementation is to monitor how many employees at trusts are logging in to the FDP’s apps. A copy of the average weekly usage data for the twelve months to June this year, obtained under Freedom of Information legislation, shows that many of the trusts classified as ‘live’ are hardly using the platform. Fifty-two of the trusts that are supposedly ‘live’ did not use a single FDP app in the twelve months to June. Across the eight apps for which the NHS has published data, usage is very low. Four were used by ten trusts or fewer over the year, including Cancer 360, which Keir Starmer described last year as employing ‘groundbreaking new technology’ that would ‘slash treatment delays across the NHS’. This app is meant to speed up diagnoses by bringing together information from cancer tests, appointments and treatments ‘so clinicians can prioritise those most in need and see patients quicker’, according to a government press release. Only six trusts used it between its official launch last September and June this year.

Some of the FDP’s apps are used more widely. The most popular, the Inpatients Care Co-ordination Solution, is intended to cut waiting lists for non-emergency surgery. It sucks up data from different IT systems within a single hospital to identify if and when additional operations can be performed. But even this tool has only been logged into at 53 trusts over the last twelve months, a quarter of the total. Kanthan Theivendran, a consultant orthopaedic surgeon at Sandwell and West Birmingham Hospitals NHS Trust, started using the IPCCS last year but stopped because he couldn’t edit or update the data. ‘I can’t do anything with the data apart from view it,’ he said. ‘It’s just a waste of time.’

Some organisations have refused to switch over to the FDP, seeing it as a downgrade on their existing systems. In private correspondence, Leeds Teaching Hospitals NHS Trust, one of the largest and busiest acute hospital trusts in Europe, told NHS England that it would ‘lose functionality rather than gain it’ if it adopted the FDP. Under the terms of the FDP contract, GP data isn’t automatically uploaded to the system, in a concession to campaigners concerned with data privacy. This means that some healthcare providers have bespoke data systems that are far more extensive than Palantir’s platform. ‘We can match anything that the Palantir FDP can do,’ NHS Greater Manchester’s chief data officer, Matt Hennessey, told the Health and Social Care Select Committee last month. Over the last six years, Greater Manchester has built a bespoke system using Microsoft Azure and software from smaller specialist firms such as Matillion and Snowflake. Hennessey told the committee that his team had also ‘spent a lot of time working to build public trust, working to get the social licence and public legitimacy’, enabling it to link up its health data with other records. This includes local GP data and adult social care data held by local authorities. For these reasons, NHS Greater Manchester says its ‘local capability exceeds anything the FDP currently offers’.

Palantir’s head of UK corporate affairs, Ben Mascall, told us that ‘Palantir software has helped to deliver 110,000 additional operations to date, a 15 per cent reduction in discharge delays and a 6.8 per cent increase in patients finding out whether they need cancer treatment within 28 days.’ These claims – which also appear on NHS England’s website – have been repeated by ministers, officials and the media. But according to data obtained by the campaign group Foxglove, in the year after adopting the platform, 13 of the 41 NHS trusts that use the FDP to schedule operations carried out 9000 fewer procedures between them than they had the previous year. And NHS England acknowledged in a methodology statement last month that any improvements in performance cannot automatically be attributed to Palantir software: ‘other variables have not been controlled for,’ and therefore ‘conclusions about cause and effect’ cannot be drawn. The UK Statistics Authority is now investigating NHS England’s use of statistics to promote the FDP. NHS England did not respond to our specific questions about the FDP, but a spokesman told us that ‘NHS England has always been clear that the published benefits are a comparison of performance before and after the adoption of the platform.’

Reporting by the Financial Times has shown that most of the supposed benefits of the FDP across the NHS rely on improvements at a very small number of trusts. For example, 84 per cent of the 217,000 outpatients removed from waiting lists across the 16 trusts that used a special FDP app were in the care of a single trust, Chelsea and Westminster. The trust is often hailed as an exemplar of good practice, but is hardly typical. For one thing, it has strong political ties to Palantir. From 2022 the co-chair of the trust (and of three other hospital trusts in London) was Matthew Swindells, a former health policy adviser at Number 10, who was hired by Global Counsel in 2019, months after stepping down as deputy chief executive of NHS England. Board papers for Chelsea and Westminster show that Swindells was ‘to be excluded from any decision-making in relation to Palantir’. He was forced to resign earlier this year after the FT revealed that he had privately urged colleagues to add GP patient data to the FDP in a bid to demonstrate its effectiveness. Chelsea and Westminster has been given an extra £3 million by the Department of Health to help implement the FDP, more than any other trust. Such additional costs, as well as the £8.5 million given to KPMG in 2024 to promote the FDP across the NHS, mean that the true price of the Palantir contract is likely to be far higher than £330 million.

NHS leaders tend to dismiss reservations about Palantir’s products. In June 2025, Tang held an event in Parliament ‘in conjunction with Palantir’ at which they ‘showcased the impact and effectiveness of our FDP’ to 25 parliamentarians including ‘multiple Labour MPs and peers’, according to documents obtained under FOI. Tang told a conference last October that the ‘egos’ of hospital bosses are the biggest obstacle to adopting new technology. Behind the scenes, disputes about whether or not to use Palantir are becoming increasingly bitter. ‘Anyone raising concerns about Palantir’s ethics and ideology are silenced and batted off with responses like “I don’t care what happens on the other side of the world,”’ an NHS data analyst said. ‘It has become a tribal power struggle at senior level. Analysts and developers are being told to aggressively press ahead with deployment, despite the fact we are giving very honest feedback that the FDP in its current state cannot do what the organisation needs.’ She added that ‘focusing on trying to make the FDP work for us is a real distraction from doing analytical work that could positively impact patient care.’ Some NHS staff have begun silently boycotting the FDP.

In private, senior NHS executives acknowledge that there are problems with Palantir’s technology – not least its deep unpopularity with data analysts. A briefing for senior NHS leaders in February complained that the FDP is ‘slow and clunky’. Processing data queries is ‘reported as taking four to five minutes compared to thirty seconds’ on existing systems. An NHS analyst who uses the FDP daily told us it’s often much slower than that. ‘I’d be happy with four or five minutes,’ he said, ‘but usually I’m waiting closer to ten.’ The briefing mentioned other difficulties. ‘Learning how to use FDP is a steep learning curve,’ it said. For NHS analysts ‘the user experience is poor’ and ‘there isn’t a clear learning pathway … which makes it harder to know where to start.’ Its dashboards – championed by NHS bosses during the pandemic – can’t ‘make simple formatting changes to how charts look’. One source who uses the FDP every day told us they often had to wait twenty minutes for a dashboard to load up with new data only for the system to crash. A Palantir spokesman responded the FDP could not be compared ‘like for like’ with other systems because of the ‘additional security’ embedded in it.

The FDP rollout has been marked by secrecy. FOI requests about Palantir’s NHS work are subject to delays and obfuscation. Democracy for Sale is taking a case to the Information Tribunal after the Department of Health refused for more than a year to release briefings prepared for ministers about the FDP. When campaigners at the Good Law Project requested a copy of the FDP contract they were sent four hundred pages of redactions. The political communications agency Topham Guerin was hired to run a campaign to counter the Good Law Project’s supposed ‘misinformation’ and ‘soothe anxieties’ about Palantir. The plan was to approach influencers on Instagram and ask them to post pro-Palantir messages in exchange for payment. Prospective participants were told to ‘keep the brand confidential and not tag Palantir’ in their posts. The scheme fell apart when Julia Patterson, who writes a popular Substack about the NHS, forwarded the invitation to the Good Law Project, which published it.

Talking to people inside Palantir is extremely difficult. Staff generally have to sign non-disclosure agreements. Reviews on the platform Glassdoor by current and former employees routinely cite salaries as a ‘pro’ (interns make £60,000 a year) and politics as a ‘con’. ‘“Flat” hierarchy means that power ends up being political, and it takes some Kremlinology to understand who is in control,’ someone employed as a deployment strategist wrote.

Concerns about privacy and Palantir’s access to data remain. The under-secretary of state for health innovation and safety, Preet Kaur Gill, had to apologise last month after NHS England failed to disclose that some Palantir staff would have ‘unlimited access’ to identifiable patient information while working on part of the FDP, the National Data Integration Tenant. In a letter to the FT, Denis Pereira Gray, a former chairman of the Academy of Medical Royal Colleges, has described the decision to give Palantir staff access to this data as ‘a historic turning point’: ‘For decades in Britain the principles governing disclosure of identifiable patient data have been that it should be either to someone involved in the patient’s clinical care or that the patient had given informed consent.’ An NHS England spokesman said that ‘the NHS has strict policies in place for managing access to patient data and carries out regular audits to ensure compliance,’ while Mascall, Palantir’s head of UK corporate affairs, told us the firm is merely a software provider: ‘How that software is used is controlled by the NHS trusts who use it, with data – legally and contractually – only able to be processed strictly in accordance with their instructions.’

The rollout has not been a total failure. Experts told us that over the past decade, NHS trusts have largely been left to develop their own data-sharing and analysis systems. Some, including those in Kent, Dorset and Leeds, did this very effectively. Others lagged behind, and these are the trusts which have adopted the FDP. As some trusts make greater use of the platform, more data flows onto Palantir’s platform, leaving the NHS more reliant on the firm. Mosley has grander ambitions: in a submission to the Covid Inquiry, he proposed building a ‘common operating system’ that would draw together data from ‘across the local and central government, healthcare and other bodies of national strategic importance’.

The FDP’s problems are fundamentally structural. Foundry is proprietary software: the NHS can license and use it, but can’t adapt it beyond the terms set with Palantir. When problems are identified, data engineers and analysts have to wait for Palantir staff to make changes. On an internal online forum called ‘FDP builders network’, one NHS technician recently flagged a problem with Palantir’s code that was creating ‘a huge amount of work’. He asked Palantir to fix it, but no one answered. Four days later he posted a gif of tumbleweed. As a week went by, others pitched in with gifs: a bemused Mr Bean standing in a field, a dog staring out a window.

If Palantir’s contract were to be terminated, NHS staff would lose access to everything there. ‘All the work we do is on the platform,’ one data analyst told us. ‘It’s not stored in files that can be exported elsewhere, and everything is done using their proprietary software.’ Even supporters of Palantir’s role in the NHS admit that this ‘vendor lock-in’, as it’s known, is problematic. ‘What matters most is that there are competitors to Palantir because what I’m most worried about is the commercial question, which is provider capture,’ a former British health secretary who worked closely with Palantir in office told us. ‘I can’t wait until Apple or Microsoft – or even better, a British start-up – is able to do what they can do and compete. My worry about the Federated Data Platform wasn’t Palantir winning contracts, it was Palantir winning all the contracts.’

Many of those we spoke to in the NHS went much further. For them, the FDP – and the decision to hire Palantir – was the wrong answer to a serious problem: how to connect Britain’s health data. ‘It’s a massive wasted opportunity, because these people have gone off and bought something without asking the people who know how this stuff works,’ one NHS analytics chief told us. ‘You could have asked us: “What do you think we should spend all this money on?” We’d have had a load of good requirements. All quite boring technical stuff: single data model, open source. Instead we got a one-size-fits-all system from Palantir that just doesn’t work for the NHS.’ ‘We keep repeating the same mistake,’ Joe McDonald, the NHS IT expert, told us. ‘We fix the problem the politicians are interested in, which may not be the problem that the NHS is interested in. But the minister likes it and the sales staff from these great big mega corporations get to the minister and he buys it. And then he’s gone before long, and the NHS has to do something with what’s left.’

The turmoil at the top of British politics has had little effect on Palantir’s capacity to win government contracts. Starmer’s Labour was enthralled by Silicon Valley and its promises of AI-driven growth. In June 2024, a month before the general election, the then shadow health secretary, Wes Streeting, said that when it came to the take-up of the FDP he would ‘go further, faster’. In office, Streeting described Palantir as ‘absolutely critical to the future of the NHS’; he resigned as health secretary in May.

In Mandelson, Palantir had a champion in the new administration. In February 2025, shortly after he became ambassador to the United States, he accompanied Starmer on a state visit to Washington DC. After a meeting with Trump at the White House, the embassy arranged another stop on Starmer’s itinerary: a visit to a Palantir showroom. Karp and Mosley were there to meet the prime minister, flanked by men in US military uniform. For 45 minutes the Palantir executives talked to Starmer about defence spending. Number 10 refused to say whether Starmer knew that Palantir was a client of Global Counsel, in which Mandelson still held a stake of around a fifth at the time. The meeting was never declared on transparency registers. In May 2025, Mandelson joined Mosley at the Hill and Valley Forum in Washington. A couple of months later, Mandelson wrote to Morgan McSweeney, then Starmer’s chief of staff, saying that the ‘celebrated techie’ Peter Thiel would be visiting London and asking whether ‘the PM would like to meet him’. That September, Karp signed a strategic partnership with the Ministry of Defence, with the government agreeing to spend up to £750 million on Palantir’s defence technology over the next five years. This was followed three months later by a £240 million contract to supply ‘data analytics capabilities supporting critical strategic, tactical and live operational decision-making’ to the armed forces. The government later admitted that the UK military was now effectively locked into Palantir’s software. Changing supplier, the MoD said, ‘would involve rebuild of the underlying data analytics architecture’.

Global Counsel’s collapse doesn’t seem to have damaged Palantir’s relationship with the British establishment. It has recently been working with FTI Consulting, another lobbying firm with strong connections to the top of the Labour Party: the former health secretary Patricia Hewitt is a senior advisor. Palantir has also intensified its PR efforts. The company’s logo may be missing from its Soho offices, but it’s regularly at the top of Politico’s influential London Playbook newsletter. Sections of Britain’s right-wing press have adopted Palantir as a culture war cause. The Times has run numerous op-eds from Mosley and a slew of sympathetic pieces. ‘Real NHS delivery means sticking with Palantir,’ the columnist Emma Duncan wrote in June. (Rupert Murdoch was an early investor in Palantir. It’s not known whether he still has a stake.)

Meanwhile, Palantir’s founders have been expanding their influence in American politics. Karp gave $1 million to the MAGA Inc super PAC after the 2024 election and a further $1 million to Trump’s inauguration committee. Palantir is one of the companies that has donated to the funding of Trump’s new White House ballroom. Thiel is throwing money at the midterm elections this November. Such spending has paid for itself many times over. Palantir has become entrenched in some of the most important areas of the US federal state: defence, immigration enforcement, data integration. In its most recently reported quarter, the firm’s revenue from work for the US government grew 84 per cent year on year to $687 million. Its contracts include $30 million for the Immigration Lifecycle Operating System, which consolidates biometric and federal data into a single platform to give ICE ‘near real-time visibility’ of people who are what it calls ‘self-deporting’.

But opposition to Palantir is starting to bite. New York City’s public hospital system, the Dutch Ministry of Defence, the French intelligence service and the Danish government have all either not renewed Palantir contracts or are actively looking for new suppliers. In June Palantir lost all but one of 23 counts in a lawsuit against Republik, an investigative website which had reported that Switzerland’s army had raised concerns that confidential military data could be passed on to US intelligence agencies if Palantir were given defence contracts.

In the UK, too, it is under increasing pressure. According to internal documents we obtained from a feedback session for data analysts last year, ‘the biggest concern with the [FDP] plans is around Palantir’s ethics and the resulting lack of trust from patients.’ In April, two petitions calling on British ministers to end public contracts with Palantir attracted more than 230,000 signatures. A month later, Sadiq Khan blocked the Metropolitan Police’s proposed £50 million deal with Palantir to automate intelligence analysis in criminal investigations, on the grounds that the contract breached procurement rules and did not ‘adequately demonstrate value for money’. (Palantir is suing the Mayor’s Office for Policing and Crime.) In June, the Science, Innovation and Technology Select Committee published a highly critical report warning that Palantir had expanded its presence across the British public sector despite a ‘clear mismatch with UK values’. The report urged the government to exercise a ‘break clause’ that would allow Britain to end the FDP contract in February 2027. Earlier this month the Health and Social Care Select Committee also called for the contract to be ended next year, citing ‘serious mistrust of the Palantir-created FDP amongst both the general public and the medical profession’. Last year, Streeting announced that NHS England would be abolished to ‘cut bureaucracy’. Next spring it will merge with the Department of Health. What this means for the future of the FDP remains unclear. It is equally unclear what consequences the accession of Andy Burnham, a former health secretary, will have.

It’s not as if there aren’t alternatives to Palantir. Estonia has developed a programme called X-Road which enables the secure exchange of data across its public and private sectors. Open-source and free to use, it has been licensed for health data in Finland and Germany, while around twenty countries including Japan and Brazil use it for other public services. In the UK, the government has moved its Homes for Ukraine website in-house after deciding not to renew its deal with Palantir. The new model is ‘already saving the Ministry of Housing, Communities and Local Government millions of pounds a year in running costs’, according to a government blog post. Local authorities using the new platform say it is ‘easier to navigate than the previous system’ and ‘safeguarding checks are easier to complete’. In May, however, the NHS closed all its open-source repositories due to fears of AI-supported hacking. (Security experts quoted in the New Scientist described the move as unnecessary and counterproductive.)

For all Thiel and Karp’s boasts of omniscience, AI-fuelled warfare and civilisational conflict, Palantir is fundamentally a software business. It doesn’t have a foundational AI model, like OpenAI or Anthropic, or build rockets or surveillance drones, like Elon Musk. What its leadership has grasped is that in an age of revanchist nationalism, there is an advantage to exaggerating one’s own threat. As the technology writer James Ball has argued, the company’s self-mythology obscures ‘what is ultimately a rather mundane product’.

Palantir – and its mundane product – is increasingly embedded across the British state. Last year Mosley was given a seat on the MoD’s industrial joint council, a forum where ministers, military officers and suppliers meet to discuss procurement. In February he presented a report to the council, co-authored by Palantir, arguing that the barriers for tech companies to win defence contracts should be lowered. His recommendations included ‘single-stage’ approvals and ‘devolving’ approval authority to more junior officials. A senior defence official welcomed Mosley’s suggestions as ‘an excellent example of collaboration at pace’.

Send Letters To:

The Editor London Review of Books, 28 Little Russell Street London, WC1A 2HNletters@lrb.co.uk Please include name, address, and a telephone number.

Footnotes

David Runciman reviewed The Contrarian in the LRB of 23 September 2021.