Written evidence submitted by medConfidential (DCG0040)

Evidence/references to SIT Committee Inquiry on Digital Government, following Oral Evidence on the 8th July

1. This evidence was written and submitted in the days after the oral evidence session, and reflects the state of public information at that time. We would hope things move on over the summer and parts may become out of date. Improvements are welcome; regressions are equally likely. The Committee covered multiple topics, and we’ll attempt to cover them in an order that makes some sense, but as with all digital government, there are overlaps in many different ways.

2. We welcome NHS England finally publishing the minutes from Check and Challenge meetings on the day after the Committee hearing, and the day after we provided the slides in Annex1 to the Committee.[1] The Department of Health in England[2] can and does move quickly when it wishes to, and glacially slowly when it prefers otherwise – minutes from the Check and Challenge group were missing for months, DPIAs are unpublished or not done at all,[3] and even when Ministers claim to want priorities which have clear public trust and wide stakeholder support that doesn’t mean delivery. Government told stakeholders they were looking at making the OpenSAFELY system for GP data available for non-covid uses way back in June 2023 when the covid-only Direction was issued, later confirming it formally in November 2023 saying “OpenSAFELY will be open to new research applicants as soon as possible in 2024.“[4] As these words are written in July 2025, NHS England has not yet made that system available for research.[5]

3. medConfidential welcomes the (limited & possibly reducing[6]) commitment in the Ten Year Plan to allow citizens to see how data about them in the Single Patient Record has been used – abuses are rife – but medConfidential lacks confidence that the Department of Health in England will keep such promises. Since the SPR is expected by DH/E to be attached to Palantir/FDP, it is of concern that NHS England’s track record is minimal in delivering on their commitments to privacy, trust and transparency. Transparency is be a political decision and without primary legislation the political decision will institutionally degrade towards a coverup. Making the existing audit logs from GP Connect and Summary Care Records (both sufficiently owned by NHSE already) visible in the NHS App would be a start.

1) Palantir

4. When Palantir and NHS England talk of “Federation”, it is tempting to think of the Star Trek universe and that benevolent Federation with Patrick Stewart in a jumpsuit quoting Shakespeare – this is how NHS England like to envisage themselves, as everyone is the hero of their own story. However, the FDP is much more like Federation in the style of early 1980s low budget BBC sci-fi Blake’s 7, which also reflects Palantir’s other customers.[7]

5. Evidence from Palantir (Q120-Q122) was that the NHS owns the uses of the system and the data flows (not inaccurate, but incomplete), yet later evidence from NHS England was that on NHS site visits they saw Palantir staff on NHS sites developing the Cancer360 tool (Q180) – the NHS may own the uses and the flows, but Palantir owns the tooling.[8] Cancer360 being a Palantir product which is tightly bound to Foundry and was built in association with NHS Trusts initially (pre-FDP) and not the national NHS England contract.[9] Evidence from both witnesses was potentially contradictory on whether a hospital using the Cancer360 tool (as NHS England is mandating) would be able to migrate away from Foundry at all, or whether Palantir would make a business decision to prevent Cancer360 and the other Palantir products running on anything other than Palantir’s technologies. We presume they would make it de facto impossible, but we have never seen an answer to the question:

Is anyone today running cancer360 on anything other than Foundry?

6. Answering Q120 “Technically, I think it is extremely low” on lock in is a technical answer. The lockin is not technical as “They own the data” (Q122), but it is worth examining “the logic” (as Mr Moseley put it). On the connections (Q122), we have heard it said by parties close to Palantir that most of “the connections" Palantir relies on for talking to NHS databases and NHS file formats is already open source code written by others (often the NHS, often publicly funded) which Palantir reuses. Neither medConfidential nor the Committee have the scope to assess how true that is for the various footprints of the NHS, but some of the technology groups in and around the NHS should be able to publish an impartial assessment (and some of them may even be independent). A carefully worded recommendation from the Committee setting out the exam question that the NHS and its ecosystem should answer would be invaluable. This is a topic for which there can and should be a set of authoritative and published answers – evidence not guesswork – and provide a steer for additional development that would be available to all.

7. Palantir’s “Common Operation System” is their brand for their UK Government vision, while in the US the same software has a different brand: “ImmigrationOS”. The difference is only in branding and the selection of features, as the “drag and drop” functionality applies between entirely different palantir products (Q125), as well as two instances of Foundry both built on top of the “canonical data model”.[10]  NHS England will pay[11] Palantir (and supposedly others[12]) to develop Palantir-only features for FDP.[13] We’ll return to the effects of this reuse in the section on genomics.

8. We do not expect that Palantir will use data to train an AI capability that has not been approved by the NHS – as things stand, we do not expect them to “go rogue” unless their US leadership decides to burn international customers in favour of only US domestic customers,[14] – but two points: they can go rogue, but equally they do not have to go rogue. NHS England will approve Palantir doing anything that is claimed to be in the interest of the NHS. Palantir’s choices for their US customers (or their middle eastern customers) suggest that Palatnir’s choices may diverge from what large numbers of NHS patients may think is appropriate. It is NHS England’s absolute and unwavering refusal to so far (publicly) consider such possibilities, which puts individual Trusts and ICBs in a position where they have a technology  imposed upon them from the centre without reassurance or guidance on that topic for their staff or patients.[15] That vacuum, imposed by NHS England by choice, which leads to individual health workers doing what the 10YP calls “shouting into the void” (page 94) and then rationally and credibly deciding their most effective course of action is to unfurl a banner outside Palantir’s London HQ. It is medConfidential’s informed assessment that that action had more material effect than all of the engagement, all of the check and challenge discussion, every engagement managed by NHS England, and all of the online petitions, all put together. Indeed, that single event appears to have had around the same number of (unique) people physically present as the (total) number of visitors to NHS England’s “FDP Engagement” website in 2024 (see figures in the Annex). We understand Palantir told the press that no one was in the office during the demo on a day just before Christmas, and we would be interested to see the response if Palantir were invited to restate that line to the standards of integrity expected of evidence to Parliament without GK Chesterton’s ambiguities.
   Photo: medAct.org / Healthworkers for a free Palestine

9. It takes politics to observe the transformation from Stalin to Mr Bean. Palantir loudly picks sides, and then the world moves on in directions that may not be compatible with those sides.[16] As with all cartoon baddies, Palantir revel in their perception and use it as a marketing ploy (Q128[17]), their marketing is aided by (US normal) political communications[18] to build their mythos, which helps increase their share price.[19] GK Chesterton observed that “evil always takes advantage of ambiguity”.[20] As a military contractor Palantir are proud[21] their tools are used to kill and maim “enemies” of Palantir’s military customers – comments that may appeal to shareholders and in industries where there are no end users with any say in decision making. In a UK NHS context,[22] such comments have a very different resonance for the close family of those supposed “enemies” living in the UK and being treated by the NHS, in some cases those maimed as “enemies” are British citizens who are then treated by the NHS – all those British citizens have a vote in how the NHS is run at election time.  Some of the rumours about Palantir have a kernel of truth, others are a series of varying inferences.[23] Both medConfidential and Trusts regularly get emails from patients intensely distressed by the entire notion of Palantir delivering the FDP, with a range of concerns spread across the spectrum of credibility. However, all patient concerns should be heard and engaged with by NHS care providers - NHS England leaves Trusts and ICBs with nothing. NHS England has decided what is best for the NHS in England and that was that - patients be damned. When NHSE expresses something as similarly tone deaf and obstinate, as they often do, we muse “maybe that’s why you are getting abolished” (and partially, it is). However, as the conclusion to Chapter 6 of the 10YP plan says, the institutions of the NHS will leave patients “shouting into a void” and the institution will do whatever it likes. Palantir can only be used by entities who do not give any regard to the political reputation Palantir cultivated for their own stock market and marketing gain, and the previous government decided that was them, and extended the contract until after this Parliament must end. Palantir would like their infrastructure a la “zero to one”, but:

if Palantir had, via an FDP (or an SPR) equivalent, serviced hospitals in Gaza on October 6th 2023, what would Palantir have done on October 8th? Which customer would Palantir have chosen to betray – or were the comments only marketing hype feasible only when a decision did not have to be made?
What if Palantir serviced an FDP/SPR equivalent in Greenland? Or Canada?

Single Patient Record and the Federated Data Platform / Palantir

10. Since FDP has become a toxic brand, NHS England has created a new one which it claims is different – the Single Patient Record. It is clear from NHS England’s own words that SPR and FDP are expected, by NHS England, to be the same thing. We understand the official position of some parts of NHS England is that this is that nothing has yet been decided (avoiding this disclosure is not why the 10YP delivery chapter got pulled, but it wont have helped). The Request for Information from Suppliers may have been little more than a fishing expedition.

11. The new Review of Patient Safety, published the week after the 10 Year Plan and written by the new Chair of NHS England, says all the safety parts of Single Patient Record will be delivered by the Federated Data Platform itself.

12. In the bottom right of the 10 year plan on page 48, it says that maternity services will be the first adopter of the Single Patient Record (SPR).

"we will roll out the SPR in maternity care first. We know from both evidence and our engagement that experiences of maternity care are not good enough, and that mothers’ preferences are frequently ignored. The SPR will ensure maternity teams have all the information they need about previous consultations, medical history and stated preferences, helping them provide genuinely high-quality, personalised care

13. What is missing from the 10 Year Plan itself was included in a DHSC press release on the 30th June the week before, which talked about the maternity care project in more detail including where it would be hosted:

"By helping transform patient care, this initiative forms a key part of the government's Plan for Change. It is built on the NHS Federated Data Platform, which allows healthcare staff to securely access the information they need in one place. That means less paperwork and manual inspections for staff, and more time caring for patients.”

https://www.gov.uk/government/news/world-first-ai-system-to-warn-of-nhs-patient-safety-concerns

14. The has been concern[24] that the 10YP makes a big deal out of genomics, while recognising much of it is mere fog. The good news is, that if a baby is born tomorrow with a rare genetic condition they can get the genetic tests already, or if a loved one is diagnosed with cancer tomorrow, the genomic sequencing promised for 2035 will be available for them today. Indeed, the 10YP is clear that if a baby is sequenced at birth,[25] and gets cancer years or decades later, they will for clinical reasons have their entire genome resequenced so that the tumor and the person are sequenced on the same machine at the same time, as they do today. By then the cost of sequencing will likely be lower than the cost of keeping the data around (especially when the consequences of catastrophic loss are taken into account). There are many medical and genomic consequences of sequencing which we will cover in a future specialist briefing,[26] while the remainder of this will cover the Palantir interactions. medConfidential were asked to write a note on how to do genetic screening in a Palantir/FDP system, and our ideas went nowhere because they did not put the database of the entire genetic makeup of the entire country on Palantir systems.[27] There are undoubtedly genetic markers that only exist in those with particular ancestry – data mining to find those hidden combinations is what Palantir was built for. The NHS will design such tools for clinical uses and Palantir may graciously offer to build and validate them at cost, because the “drag and drop” nature of Palantir’s tools and the Canonical Data Model means they will work without change for other Palantir customers who will use them for their own ends. When something is technically possible, unless technically and legally prohibited, it will end up being done. We will return to eugenics in the context of Biobank later, but the GK Chesterton quote from earlier in this document was in the context of population eugenics that Palantir’s amorality may facilitate if a customer desired it.

FDP as the implementation vehicle for the 10YP

15. The point of the FDP is to let the Secretary of State do anything they choose with the data within – and Trusts will be obliged to let them because NHS England (ie Secretary of State via the joint data policy unit) is a joint data controller for all data within the FDP.[28]

16. NHS England does not like disclosing how much Palantir costs – largely because we understand the budget has ballooned already. Some figures are public,[29] with a £25m target budget predicted for the first calendar year (Nov 23 - Oct 24). We read the figures for the first year:

-          £6m for startup costs for core platform from Dec/23-May/24 (almost all palantir+AWS+ NHSE, some procurement/post-signature contracting) before any data/users/etc migrate in

-          £2m for FDP pilots  from Mar/23-Feb/25 (migration, most costs already funded elsewhere)

-          £28m for pre-FDP-palantir/early adopter migration in 2024 (almost all Accenture, a bit of Palantir when Accenture need it)

-          £7m for development (mixture of palantir, Accenture, and some NHSE)

(These descriptions are all from medConfidential. There will be AWS fees rolled into each of those for additional resources above the base level but most of the heavy compute uses haven’t been migrated to FDP as of the time of those figures.)

17. We fully expect NHS England will refuse to confirm those figures, and in any event they are nearly a year out of date (ie they’ll all have gone up). Should NHS England wish to claim they were never accurate, we would remind NHS England not to redact documents only by placing a black box over some text. Giving powerful technologies to organisations incapable of using them well ends badly for the individuals involved when they blindly follow a process they do not understand, as Dan Davies’s great book “The Unaccountability Machine” recently showed with what Heywood Quarterly describes as “a memorable, and horrifying, story” about squirrels.[30] NHS England would similarly have just blindly followed policy into the shredder.[31]

The end of the FDP

18. The FDP will not be forever. Towards the end of the seven year term, the Department of Health in England will have a decision to make about what happens next. Palantir will have followed their commercial self-interests in attempting to burrow so deep into the NHS that they can not be migrated away from and can name their fee. It is the responsibility of the current DH/E staff to prevent today what they may be arguing in favour of in a future much more well recompensed role (Q180). The original workplan for FDP required an exit plan be developed, and we understand that was ticked off with a “someone else will figure it out if we ever need it” superficiality.

19. Lobbying aside, there is nothing unique or special about Palantir – indeed, the most unique or special thing about Palantir is why their off the shelf technology is seen as so unique or special. It is a marketing innovation not a technical one. By the end of the seven years perhaps many of the other organisations who could compete will have caught up. Generic technical competence should not be a unique selling proposition in the UK technology and consulting world – that it perhaps may be so is a consequence of other poor decisions, some of which Dr David Cleevely talks about at length and with examples in his new book.[32]

20. The current government, and Palantir and their acolytes, all argue in their own ways that AI is The Future. If they are correct, a future migration away from Palantir by having a coding AI read the NHS owned documentation and translate it into new NHS owned code for the workflows and data sources, all running on a generic installation of Apache Spark[33] (the open source software that Palantir Foundry is a branded installation of – Foundry is basically Spark with a lobbying arm). If the claims for AI are real, and NHS England really wishes to have the option to move away from Palantir at the end of the contract, then, in line with the rolling evolution where patented prescription pharmaceuticals have a clear and common roadmap to migrate towards generics, NHS England should design that migration contingency plan on an ongoing basis, and start implementation with an entirely-NHS-controlled version of the “Referral to Treatment” tool, following up with another module. It will also ensure NHS England can become and continue to be an intelligent customer.

2) FDP Check and Challenge

21. Dr Vin Diwakar gave evidence to the Committee about the FDP Check and Challenge group. Unfortunately that evidence itself has some challenges and needs checking. Answering Q175 he said:

Dr Diwakar: “We have a “check and challenge” session that brings in people like the BMA and privacy campaigners to make sure we get robust challenge from that. We are aiming to hear from the whole spectrum of views, both patients who are comfortable with data use and those who are privacy campaigners and raise often legitimate concerns.”

1. He explicitly says twice that “privacy campaigners” were invited on to the group’s membership – can he cite the “privacy campaigners” amongst that list? That is the usual euphemism used to include medConfidential, and we understand we were ‘banned’ from the group by NHSE’s decision makers and NHS England do not send us paperwork (although we do sometimes send them comments on paperwork they refuse to send us which they usually remain silent on).

2. NHS England has not published the Data Protection Impact Assessments as required - following a letter to Secretary of State,[34] they are starting to, but there are still a lot of gaps in this page:

https://www.england.nhs.uk/digitaltechnology/nhs-federated-data-platform/security-privacy/nhs-fdp-information-governance-framework/national-fdp-products/

3. The most notable gap is the lack of a DPIA for the “Privacy Enhancing Technologies” features of the FDP, where despite saying NHS England would do any required DPIAs,[35] they now admit they didn’t (on the basis that having privacy enhancing in the name meant there wouldn’t be any adverse impacts on privacy!) and are now doing it:

https://www.whatdotheyknow.com/request/publication_of_nhs_pet_dpia#incoming-3062690

4. When Dr Diwakar says (Q171) “under legislation at the moment, you cannot opt out of data sharing for clinical purposes for direct care use” he means that the only choice patients will have is their data or their life, continuing “that if you still want to opt out, you just need to Google search “Opt out,” and there is a simple, single, easy way of doing it.” How does that work for a dependent child under-13 living at home? Is there a “simple, single easy” way of doing that?  We fully expect, when FDP expands to non-direct-care uses, NHS England will use one of the many loopholes in the opt out to avoid respecting them even then, because NHS England believes the opt out should not apply to NHS England’s planning, despite the opt out being defined as “research and planning”.

5. NHS England tells the public they can "Access contact details for data protection officers in the NHS trusts using this product” - one only needs to click that link to see that is not true. We understand NHS England’s desire to keep secret the list of Trusts using FDP to shield them from questions from their staff, but NHS England sees no barrier to telling the public one thing and doing the exact opposite.

22. [this paragraph was originally written the day NHS England gave evidence, one day before NHS England finally published the minutes for the meetings in 2025 having failed to do so for 6 months]. In answering the above questions, as they did in oral evidence, NHSE will normally mention they are very proud of the engagement work they have done, and that they have an engagement website for the public. For the 21st March 2025 meeting they told the engagement group about how many people had used the site - Annex 1 contains the presentation with the confidential figures in it - slides 3 and 5 have numbers, all of which are utterly pitiful. Every person in England will have their medical notes included in FDP if the programme succeeds – NHS England can not reasonably claim a website with 230 people visiting it has meaningfully informed the public of anything.[36]

23. Nothing epitomises NHS England more than the deliberate duplicity on FDP engagement and transparency. Dr Diwakar talked about patients being able to opt out, knowing full well for every current FDP purpose NHS England does not respect opt outs for the reasons he outlined,[37] a lack of clarity about which page 94 of the 10 Year Plan is very clear:

“The NHS never has the right to keep the public in the dark. That it so often does so reflects the centralisation of power and disregard for patient voice we identified in chapter 5.“

3) The data in the future Health Data Research Service

24. Government telling patients that if they want NHS care the NHS must sell their data is a major and radical new step,[38] one this Government seems to have made without really thinking through the consequences or possibly even actively deciding to do it at all.

25. In Q171 Dr Diwakar said:

“There is a broader question of whether, in any forthcoming NHS Bill, Parliament and Ministers may wish to look more radically at things like the GDPR and the use of data for health purposes. I will not comment on it here, but it is a legitimate question—not just because the research will have benefits to patients directly, but because it will add benefits to the economy. We know it is one of our greatest assets.”

While the Department of Health and NHS England have had no discussions with medConfidential about changes to the data rules (those discussions have been promised, they have not materialised, even as texts are promised to other stakeholders within weeks). Maybe those discussions are at the Check and Challenge group we are not invited to (Q144).

26. The Health Secretary has talked about wearables on the NHS. It is Government policy[39] that if this data is to be made available in future to any doctor it must be made available to all doctors treating that patient, and the only mechanism envisaged to do so is the NHS App and Federated Data Platform. Health settings on iOS and Android devices are not effective against such compulsion – your data or your life.

27. Any data in the FDP can be sold on for research and commercial uses. Similarly, the analyses run in FDP are designed to be reused (that’s the “federated” bit) as Palantir’s witness said (Q125-126), in other Palantir instances. Now, subject to the approval of whichever organisation developed them (currently that is limited to NHS England and Palantir, but could be anyone approved by NHS England as described in Q123).

Biobank & Our Future Health & HDR

28. The culture that led to HDR and Biobank’s difficulties is identical to that which caused the recent catastrophes at the Office of National Statistics,[40] (and similar to the Turing which we come back to later) resulting in the resignations of the majority of the leadership of ONS and significant Parliamentary scrutiny.[41] ONS continued their approach from the pandemic assuming “peacetime” was the same as “wartime”, and that what was appropriate in a private company is equally appropriate in a state provided service backed by criminal sanction, forgetting that while using Uber or facebook is largely optional paying taxes or healthcare is not. ONS expects it will have to go back to each citizen and ask them to fill in a census form in 2031, but with administrative data it assumes it can take whatever data it can find and use it as it sees fit because the public have no choice – those two positions are not compatible. Similarly, while the NHS 10YP talks about patient voice via outcome/experience measures (PROMS/PREMS), NHS England is abolishing the Health Survey for England which provides independent baseline data. Administrative data is not enough, as the economic statistics have shown. It was the Boris Johnson administration which encouraged ONS and HDR to adopt a culture that made their decisions about transparency and accountability, decisions that now turn out to have been toxic folly – ONS is turning itself around, HDR and Turing refuse, Biobank remains to be seen.

29. medConfidential welcomes Vin’s description of the problems with Biobank's access process being under “old standards” and the implication that NHS England expects they will improve. We welcome Biobank’s change to preventing insurers accessing UK Biobank data, after previously justifying giving access.[42] It is generally notable that Biobank continues to insist they did nothing wrong.[43] The senior academic who lauds the Biobank application process with all its weaknesses is involved in designing the HDRS project for HDR and still insists Biobank were in the right. This will recur, with more data as Vin outlined in his confirmation that COPI reform is coming to use data much more widely with less safeguards. On the Biobank eugenicists, on Palantir, on NHS England, we note GK Chesterton’s 1922 observation: “evil always wins through the strength of its splendid dupes; and there has in all ages been a disastrous alliance between abnormal innocence and abnormal sin”. That is how Palantir and Biobank expect to survive.

30. If the Committee were to invite Biobank to give evidence about sending data to Chinese entities, we expect Biobank would say:

“... research is good, more research is better … so we give access to our data to Chinese researchers…”.

We equally expect Our Future Health would sit next to them and say:

“... research is good, more research is better … so we do not give access to our data to Chinese researchers…”.

Both positions can be defensible – it’s a complex area full of nuance – but the two arguments next to each other from adjacent (and competing) organisations using the same data access platform for the same types of data are utterly incoherent and incomprehensible, and make it look like a bait and switch to get data and then do as decision makers wish. Looking at the biobank website on the 9th July, 9 out of the 10 latest projects were from China.[44]

4) HDR/Foresight AI

31. The public were told in the NHS England’s data use registers that this data was “covid only” and there was no sub-licensing possible, which we understand was correct from NHS England’s perspective but HDR misled NHS England about what they were actually doing. NHS England say they have an investigation ongoing (with the narrowest remit possible), and BMA/RCGP have written to the ICO[45].

32. Data use can be good and safe, but that’s undermined by processes that are secretive and dishonest or untrustworthy providers. NHS England writes the rules, and then applicants try to game them. HDR’s researchers, those within the HDR cartel boundary, filled in the application forms that HDR told them to – and reasonable project researchers did what they thought was approved. Equally, it was not unreasonable for NHS England to assume that the National Institute for Health Data Science was acting in the public interest and being honest when they claimed to be doing covid-only work – the problem is the evidence now shows they weren’t[46] – HDR took an approach to transparency and integrity around covid that seems inspired by Baroness Mone.[47]

33. Upon questions being asked, NHS England made two decisions, the first is the decision to pause the project itself and alone (Q173) and run an investigation, but made a second decision to have all similar projects currently going through HDR’s internal application process go into the normal NHS England review process. When NHS England’s independent Advisory Group on Data looked at those projects, (item 5.1 of these minutes - 5.2,5.3 and 9 are about hiding other data uses from the public), AGD could not support the majority of HDR’s decisions. These are indications of a systemic problem that HDR and NHS England are ignoring and not investigating. HDR will likely insist (again) that HDR have addressed all questions from medConfidential, but saying they’ve addressed the question does not mean they have given answers. At the time of writing we’ve seen nothing from HDR responding to the AGD assessment.

34. We opened this evidence by noting the delays in the OpenSAFELY system and the lack of resolution in the 2 years since that work started, using project funding at the University of Oxford. In comparison, HDR and FDP have almost infinite budgets[48] and endless timescales to deliver few academic outputs but surprisingly large controversy, while systems with consensus support that deliver previously unavailable data to researchers languish caught NHS England’s institutional malaise – it entirely explains the HDR leadership decision to lie to NHS England and deceive the public over the Foresight AI project and the 100+ other similar projects that have had no scrutiny as yet.[49] We’re not arguing that HDR’s actions were irrational, we are arguing their actions were unlawful, but either way they have delivered nothing useful – all the data of the NHS, all of these projects,[50] no meaningful oversight, utter freedom with unlimited scope, and no meaningful discoveries.[51]

35. HDR’s institutional denial is analogous to that of the Turing and to that of the ONS. Turing may or may not have been retrieved from institutional denial but currently has no health remit.[52] With many of the “innovation” institutions around health research are currently in flux – Turing moving away from health,[53] alongside the 10YP and plans for patient data, the HDRS, a reset of HDR, would together give opportunity for a better foundation for the next decade of health research. The bottom line for patients should be really simple: no patient who has opted out of their data being used in research and planning should have their data used in research (or planning) they don’t want to be involved in.

5) Digital Government

36. Data always leaks. We write these words as the news emerges that the Ministry of Defence lost a spreadsheet of everyone who had helped our nation in Afghanistan.[54] We recognise that HMG does the right thing eventually, although it may try doing everything else imaginable first and advocates for the worst policies never truly believe they were wrong. It was the 2007 tax discs debacle that destroyed the credibility of the early 2000s Identity Cards scheme, but some ideas keep returning wrapped in some new technology only to die for equivalent reasons. It is the culture and incentive of tech startups that the old reasons are wrong, and politicians like to support people who agree with them.

37. Belief of civil servants and NHS officials (exemplified in oral evidence by Dr Diwakar) that if the data exists somewhere, it must be available for any purpose the State thinks proper,[55] and that if data isn’t available then it must be made available.[56] That was the ethos behind the HDR/Sudlow Review into use of data, it is the ethos behind FDP and institutional use of Palantir. Palantir isn’t magic,[57] but Palantir excels at selling military grade technologies to companies and governments; governments which may then use them against their own citizens when even well meaning civil servants given a public task (or Palantir staff given their OKRs) are incentivised to draw no distinction between “peacetime” and “wartime” to use Dame Kate Bingham’s framing of covid.[58] The civil servant in the American South prosecuting teenagers for missing their period believes in their public task just as much as any civil servant on Marsham Street or in DHSC. They all want to make their jobs easier and suppliers are eager to make money.

38. The Science, Innovation and Technology Committee of the House of Commons has a unique remit to look at the boundary of discovery, a boundary that is constantly evolving. Science, Innovation and Technology can do wonderful things and AI is assisting with that, but there are consequences:
1. By the end of this decade it is likely that many cancers will be curable in your cat, at their normal vet, meaning it is politically untenable that the NHS will not similarly cure them in your loved ones; a topic the 10 Year Plan covers briefly in two paragraphs on page 71. The political consequences and the NHS service redesign questions are profound (and the availability of double figure percentage of the NHS budget that goes on cancer and consequences of cancer may make them comparatively easy to deliver). Companies know from US colleagues that people will pay any amount necessary for health care, and health data is uniquely identifying as denial often means death.
2. LLMs and chatbots are like kind elderly committee witnesses, they may want to be helpful but sometimes need careful nudging to answer the question you wanted them to answer and not go off on tangent, which in an NHS context can have physical consequences.[59] We are currently at the point of chatbots where privacy-preserving AI chatbots are good enough that if you’ve have had a bunch of tests that haven’t figured your condition out, you as a patient can run your own records through a privacy preserving chatbot which keeps no trace of your data or query, which can ask you a set of questions that you’d never think to talk with your doctor about, and your doctor would not know to ask you. The NHS is a long way from offering that, simply due to compute requirements, but you can run it on your mobile phone using technology available to you today. That same technology could also scan all recent correspondence and check that any recommended followup is actually being followed up.[60] The power for AI in the NHS, and in digital government, comes from doing new things for earlier and better care, and not focussing on saving 20 seconds per patient in metrics where Trust managers will end up berating doctors for speaking too few words per minute. What gets measured gets managed and some things are easier than the better things.
3. Proteomics has been “The Future” since the 1990s, and UK research have long required proteomes be published openly on the internet for research use,[61] despite research advances in the interim allowing anyone to go from a proteome via the proteins to the DNA sequences that would have created them – proteomes are now as identifiable as genomes, but the practice of publication of proteomes continues in a way we do not accept for DNA.[62]
4. The same technical commitment that allows Palantir to drag and drop tools across their product range (Q125), and across tenancies in the Federation (Q169/Q170), means when one Palantir-using location builds a tool using sensors on devices running the NHS app, an App which follows DH/E policy to upload all sensor readings to Palantir/FDP, then a module running in Palantir could work out when someone is pregnant,[63] with FDP knowing before the woman does. There are reasons different entities may build those tools for the NHS, and FDP would by design, mean those algorithms  clinically assured by the NHS will be entirely compatible with platforms sold to Palantir’s US customers, where women have much less control over their own bodies than Palantir wish their government customers to take.

39. Mr Streeting’s vision for the NHS is based upon someone like him being in his job forever. Should it be like his predecessor Mr Lansley it would be very different; and his successor may come from the whole range of political hues. Digital Government is entirely dependent upon the Government of the day, who may care neither for GK Chesterton’s observations on eugenics, nor for his fence.

16 July 2025

Annex: March 2025 slide pack on the https://fdp.england.nhs.uk public engagement site

The figures but not the graph from these slides are now in the public domain via the minutes of the meeting published after the oral evidence session: https://www.england.nhs.uk/long-read/federated-data-platform-check-and-challenge-group-minutes-and-action-notes-21-march-2025/

20