Whose evidence is the National Data Guardian relying on over the FDP?

On 9 October the National Data Guardian, Dr Nicola Byrne, wrote to the Minister for Health Innovation advising that public and professional trust should carry substantial weight in decisions about the Federated Data Platform. She says she strongly supports the FDP programme and does not recommend a particular contractual outcome, and writes an annex outlining her evidence. HSJ reported that Palantir accused her of being captured by campaign groups. To me, the real question is where the evidence in her letter came from, and whether she was playing into campaigners' hands by using it, intentionally or not. I went through the annex. Every category of evidence in it follows a manipulation technique designed to bypass facts and appeal to instinct. Each one of these techniques has been studied and named by scholars modern and ancient, and this article maps her evidence against them.

Cherry picking

Cherry-picking is the selection of evidence that supports a predetermined conclusion while leaving out what contradicts it. Byrne has used this several times in her letter.

Firstly, the letter gives prominence to the more than 100 NHS data and digital staff who signed a letter against the Palantir contract. That is just 1%, considering an estimated 10,000 analysts work in the NHS. If anything, this figure tells the opposite story to the one Byrne has told.

Secondly, she questions whether the FDP's benefits have been proved. The Office for Statistics Regulation reviewed the FDP benefits statistics, required stronger caveats around cause and effect, and did not require withdrawal of the published figures. The letter cites the review and presents the FDP benefits as thereby contested, where in fact there are good benefit reports available in the public domain such as the Imperial study at Chelsea and Westminster published in BMJ Health Care Informatics.

Thirdly, the same selectivity runs through her letter's treatment of other suppliers. Billions have been spent on electronic patient records without anyone demanding proven benefits. The letter raises national data sovereignty as a concern, but this is a test no US supplier can pass. Applied consistently, it would disqualify Oracle, Epic, Microsoft and AWS alongside Palantir. Microsoft Copilot is being rolled out to half a million staff without a sovereignty review. Applied to one supplier, it becomes a criterion designed to produce a single outcome. Byrne's letter notes that other US suppliers have not faced similar scrutiny, but never asks why.

Advocacy polling

The American Association for Public Opinion Research defines advocacy polling as campaigning disguised as research: a small number of questions about a single issue, designed to persuade rather than measure opinion.

The letter cites a YouGov survey of 2,097 adults, commissioned by Ekō. Byrne does not mention that Ekō is a campaign organisation. The first question asks whether it is acceptable for US firms to handle patient data on behalf of the NHS. This is a priming question, designed to frame what follows around loss of sovereignty and people's fears of large corporations. It gave respondents no context about how the NHS already uses US software for the electronic staff record, email, office tools and the electronic patient records where clinicians read your diagnosis. Oracle, Epic, MEDITECH and InterSystems are all American, to name a few. Taken at face value, the survey tells us the public supports none of this.

The second question asks whether contracting a US firm would make people think worse of a political party, which tells you the poll was designed for lobbying rather than understanding.

Palantir only appears in the third question, after the first two have set the frame and the respondent has already been primed to find US firms handling NHS data unacceptable. Here the "evidence" is generated: 65% of people think the government should not continue with the current FDP contract. By the time they have answered this they have been given no context, no benefits of US technology, and a frame that puts them in a defensive mindset. Even then, 28% said they did not know, and YouGov's brand tracker shows only 38% of the UK public have ever heard of Palantir.

Circular evidence

Circular reporting is a recognised technique in which information from a single source is passed through intermediaries until it appears to come from multiple independent sources. In a campaign context, the group generates the media coverage and public correspondence, then cites the volume of that coverage as independent proof of public concern.

The letter draws on Understanding Patient Data's analysis, which found 243 articles on Palantir and the FDP, 84% negative, and on reports of large volumes of patient correspondence. Notably, the Financial Times has given the programme's figures more sustained scrutiny than any other NHS programme receives, and campaigners amplified that coverage to cast doubt on its value. NHS trade publications HSJ and Digital Health have maintained a consistently critical editorial line on the FDP, and that coverage feeds the same cycle. Campaigners such as Good Law Project run sustained social media campaigns that amplify each piece of coverage.

Coverage and correspondence measure how active a campaign has been. A well-organised campaign will generate plenty of both. Neither tells you how much the public trusts the arrangements.

Repetition as corroboration

The illusory truth effect is a well-replicated finding in cognitive psychology: repeating a claim makes it seem more credible, even when nothing new has been added. Presenting the same recommendation from two sources as independent corroboration exploits this effect.

The SIT Committee's Rewiring the State is a report about digital government across the whole public sector. Palantir holds contracts in policing, defence and other departments, yet the only contract the report recommends terminating is the NHS FDP, which it describes as "an unacceptable point of weakness". That happens to be the one contract Martin Wrigley MP, one of the SIT Committee members, has campaigned against.

The Health and Social Care Committee then wrote to the government echoing the same recommendation. Byrne's letter presents these as two separate pieces of evidence, but in substance they are one argument repeated.

Martin Wrigley MP, the most active backbench campaigner against Palantir, sat on both panels.

Byrne also cites her own previous statements as part of the evidence base: a November 2022 blog, an August 2023 blog and a June 2026 statement, each warning about trust. She now presents the pattern of her earlier warnings as evidence that the trust problem she predicted has materialised. That is one person's view, counted three times, as though it were accumulating independent evidence.

False context

The Council of Europe's framework on information disorder defines false context as genuine information presented with misleading framing.

The letter describes rising opt-out rates as a significant concern, while reporting the minister's response that there was no evidence of significant impact. The contract was awarded in November 2023. The rises came over two years later, coinciding with campaign activity and media coverage that peaked at 71 articles in June 2026, the month of the NDIT story, which turned on an error in NHS England's data governance documentation.

The opt-outs track the campaign activity, not the contract. Those opt-outs achieved nothing for the people who made them. Anyone who opted out to keep their record away from Palantir removed it from research and planning and left it exactly where it was in the FDP.

Campaigners helped create that confusion through videos and campaigns implying that opting out would keep data away from Palantir. The letter cites opt-outs without any attribution to the campaign activity that drove the rise.

Appeal to emotion at the expense of fact

Scholars have documented this tactic since Aristotle distinguished between logos (appeal to evidence) and pathos (appeal to emotion) in the fourth century BCE. Modern rhetorical scholars such as Roberts-Miller classify it as part of a demagogic repertoire: replacing evidence-based debate with emotional narrative, so that facts become irrelevant to the audience.

The letter leans on three terms: trust, values and ethics. Each carries both a factual test and an emotional pull. The factual test can be measured, checked and passed or failed. The emotional pull works without evidence and cannot be tested at all. In each case, the letter ignores the test and relies on the pull.

Trust, for a supplier handling NHS data, means keeping data secure, processing it lawfully and using it only as the NHS instructs. No data has been stolen, lost or used for Palantir's purposes. The one incident the campaign has seized on, the NDIT access issue in June, was a documentation error by NHS England for which the minister apologised to MPs. No data was improperly accessed. Yet the letter says none of this addresses her concerns. The National Data Guardian exists to check whether data is kept safe and used properly. By her measure, it has been. She has chosen to look past that.

Values are defined in the NHS Constitution. They describe how the NHS treats patients and staff. Suppliers are not bound by them. No supplier has ever been assessed against them, and no procurement framework requires it. If values were being applied as a principled test, TPP would be first in line. Its CEO Frank Hester made comments about Diane Abbott that the Prime Minister called racist and wrong. The company was responsible for a data breach that exposed 150,000 patients' records and holds 60 million UK medical records under contracts worth more than £400 million. There was briefly a campaign. No committee recommended ending the contract. No National Data Guardian wrote to the minister.

Ethics, in this debate, means Israel, Gaza, US immigration enforcement and the views of Palantir's co-founder Peter Thiel. Byrne's annex cites Thiel's description of the British attachment to the NHS as "Stockholm syndrome" and his suggestion that "you just rip the whole thing from the ground and start over". Under the contract, Palantir is a data processor acting on the NHS's instructions. What its chairman thinks about the NHS has no bearing on how data is processed, any more than a car manufacturer's chairman's views on public transport affect how their vehicles are serviced. Many technology companies support the governments of the United States and Israel without calls for them to be removed from the NHS. These are emotional triggers presented as evidence in a letter that is supposed to rest on fact.

What the letter weighs instead is a general sense of unease that no one can measure, and whose only available measures are the campaign outputs described above.

My interpretation

The techniques documented above are not obscure. Cherry-picking, advocacy polling, circular evidence, false context and appeal to emotion have been studied for centuries and are catalogued in frameworks that any policy professional can find.

And yet Byrne describes public opposition as "striking" and says the evidence has "not assuaged public concerns". Both characterisations rest entirely on the campaign evidence described above.

Each of the National Data Guardian's three recommendations follows one of the patterns described in this article. The first relies on a definition of trust so vague that it can only be measured by campaign outputs. The second asks for clear, objective evidence of benefits, applied to the FDP at a level of scrutiny that surpasses even clinical service performance and quality data, and to no other technology contract in the NHS. The third applies a sovereignty and lock-in test to one supplier that no US supplier can pass, while leaving every other US contract untouched.

The National Data Guardian is a non-regulatory, advice-giving role whose statutory remit under the 2018 Act is to issue guidance on the processing of health and adult social care data. As a public appointee, Byrne is bound by the Nolan Principles, which require holders of public office to act on the best available evidence, impartially and on merit. Instead, every category of evidence in the annex reproduces the campaign’s output, unqualified and unchallenged, in a letter timed to land just as the break clause decision is due.

The question that would establish whether Palantir's accusations of Nicola Byrne's capture by campaign groups stand up is this: did she understand the shortcomings of the evidence she used as the basis of her recommendations to the minister?

Tom Bartlett is the founder of Bartlett Data Ltd and a former Deputy Director of Data Engineering at NHS England. He writes and advises independently on NHS data infrastructure, including the Federated Data Platform.