Last updated on May 11th, 2026 at 02:07 pm
A companion to “The Palantir Map.” If Canada is shows where dependency is being built, the other jurisdictions are the cases where we can see what happens at each stage of the build.
UPDATE 5/11: NHS Access
An internal NHS briefing note seen by the Financial Times and reported on May 11, 2026 confirms that NHS England has agreed to create an “admin” role granting external contractors, including Palantir staff, unlimited access to identifiable patient data within the National Data Integration Tenant. The NDIT sits inside the Federated Data Platform and holds patient data before it is pseudonymised and shared with downstream systems. The new access pattern replaces the previous requirement that any individual working with the NDIT apply for specific data-set access on a per-task basis. The briefing note itself recommends capping external admin access and making such permissions time-limited and subject to regular review, indicating the change is contested inside NHS England, not only outside it.
The briefing note also confirms why the access framing matters: “considerable public interest and concern about how much access to patient data Palantir/Palantir staff have.” Palantir’s position is that as a designated data processor it can only act on instructions from the data controller, and that granular access controls are overseen by the NHS. Both characterisations can be simultaneously accurate. The contested question is whether a control regime that depends on NHS oversight of contractor permissions can be operated by an NHS that is still being built into dependency on the platform those contractors are deploying.
For the staircase argument, this update changes nothing structural and sharpens one specific thing. The UK was already located between Canada and the United States, with the break clause as the live political instrument. The disclosure that contractor access to identifiable patient data is being expanded, at the same moment the Westminster Hall debate established the break clause as a live decision, converts the spring 2027 question from “is the platform delivering value for money” to “is the access architecture defensible.” That is a harder question to answer in Palantir’s favour, and it is the question that will now sit in front of the minister at the break point.
The No Palantir in the NHS coalition, for which there is no Canadian equivalent, mobilised on the disclosure within hours of publication.
ORIGINAL POST: The weekend the vendor discussion went political
Between Saturday evening and Monday afternoon, April 18 to 20, 2026, three governments engaged with Palantir in public, in three different postures.
On Sunday morning, the Prime Minister of Canada told the country that reliance on the United States had become a weakness. He invoked the War of 1812. He said the correction would require structural changes at home.
On Sunday morning, Palantir Technologies posted a summary of its CEO’s book manifesto; 22 numbered theses advocating hard power, universal conscription, the undoing of postwar pacifism in Germany and Japan, cultural hierarchy over pluralism, and an AI deterrence regime replacing the atomic one. The company prefaced it with “because we get asked a lot.” It was published under its corporate brand, on the record.
On Monday morning, Westminster Hall in the UK Parliament held a debate on the NHS Federated Data Platform. The junior health minister told MPs the £330 million contract with Palantir could end at the break clause in spring 2027. Liberal Democrat MP Martin Wrigley raised the 2019 meeting between Boris Johnson, Dominic Cummings, and Peter Thiel, a meeting for which no minutes exist, as the origin point of the UK dependency. MPs cited Palantir’s work with ICE, its contract with the Israeli military, and the political positions of Thiel and Karp. The minister said patient safety and value for money were the criteria for the break decision.
Five governments. One vendor. Three different places on the same curve.
The staircase
Palantir’s relationship with a national government is not a single event. It is a process with identifiable phases. Every country that now carries a Palantir dependency, or has declined to, can be placed on the same five-phase staircase.
Phase 1 — pandemic or crisis entry. A short-term, low-cost, often pro-bono engagement during a national emergency. Procurement discipline is relaxed because the crisis is immediate. The engagement establishes the vendor inside the institution.
Phase 2 — pre-approved vendor status. The vendor is admitted to a central procurement catalogue or framework. The political decision to allow the vendor access is made once, centrally, and does not recur.
Phase 3 — standing agreement. A multi-year ceiling is established. Individual purchases under the ceiling proceed through simplified procurement that does not attract parliamentary scrutiny.
Phase 4 — institutional lock-in. The vendor’s platform becomes the infrastructure. Multiple contracts are consolidated into single enterprise agreements. Competitive bidding is bypassed by design, not exception. The platform is classified as a program of record with protected budget status.
Phase 5 — declared ideology. The vendor publishes its political commitments under its corporate brand. Procurement after this point is procurement with notice.
Phase 5 is not a destination beyond Phase 4. It is a disclosure event that can occur at any point, but carries different weight depending on where the jurisdiction sits on the other axis. For the United States, where the infrastructure is already built, the manifesto is a retrospective framing. For the United Kingdom, where the break clause is live, it is contemporaneous context for the decision. For Canada, where the Supply Arrangement runs to 2028, it is notice that procurement choices made from here forward are different in character than choices made before April 19.
Here is where the jurisdictions stand on April 21, 2026:
The United States — what the end state looks like
The US is the reference case. It is the jurisdiction where every phase has completed, and where the vendor’s relationship to the state has reached infrastructural status.
The Army Enterprise Agreement signed July 31, 2025 consolidated 75 existing contracts, 15 prime and 60 subcontractor, into a single ten-year vehicle with a $10 billion ceiling. The service’s Chief Information Officer described it as the first of “many enterprise licensing agreements” the Army intends to sign. The agreement is not restricted to the Army; other Department of Defense components can purchase under it.
The Department of Homeland Security Blanket Purchase Agreement published in February 2026 established a five-year, $1 billion ceiling covering ICE, Customs and Border Protection, the Transportation Security Administration, FEMA, the Secret Service, and CISA. The BPA allows these agencies to procure Palantir software without competitive bidding up to the ceiling. A Palantir executive described it internally as the “next chapter of Palantir’s work across DHS.”
Maven Smart System, the battlefield AI platform Palantir operates for the Pentagon, had its contract ceiling raised from $480 million to $1.3 billion in May 2025. Deputy Secretary of Defense Steve Feinberg designated it a formal program of record in March 2026, placing it in the Future Years Defense Program as a protected budget line. More than 20,000 active users now rely on it across every US combatant command. A Pentagon official demonstrated its targeting capabilities for Middle East operations at a Palantir event.
Palantir’s US government revenue grew 66 percent year-over-year in Q4 2025, reaching $570 million in a single quarter. The company’s full-year US government revenue for 2025 was $1.855 billion, up 55 percent.
Alongside this procurement arc, the institutional proximity question has resolved in a specific direction. The Vice President of the United States, JD Vance, was hired into Peter Thiel’s Mithril Capital in 2016, launched Narya Capital in 2019 with Thiel backing, and reached the Senate in 2022 with $15 million in super PAC support from Thiel, at the time the largest single-donor contribution to a Senate race in US history. Thiel introduced Vance to Donald Trump at Mar-a-Lago in 2021 and reportedly lobbied for his selection as running mate in 2024. Thiel co-founded Palantir in 2003.
The United States is the jurisdiction where the institutional proximity is not a compliance question. It is the governance structure.
There is a tell inside the US case that is worth naming. According to reporting in the defense press, the Pentagon has begun testing AI models from Microsoft and OpenAI specifically to reduce vendor dependency on Palantir and Lockheed Martin. The largest customer of the platform is actively building redundancy against it. Allied states that are still in Phase 3 or Phase 4 are being asked to embed deeper into a vendor than the vendor’s own primary customer is willing to.
The United Kingdom: contested mid-stage
The UK sits between Canada and the United States on the staircase, and is the only jurisdiction where the political system is actively trying to pull the dependency back.
Palantir entered the UK in 2020, on a £1 initial contract during the pandemic. This is the canonical Phase 1 move. By 2023, the Federated Data Platform contract was awarded to a Palantir-led consortium including Accenture, PwC, NECS, and Carnall Farrar, worth up to £330 million over seven years with a break clause at year three. Palantir has subsequently expanded into policing in 2024, the military in 2025, and financial services in 2026 via the Financial Conduct Authority. An investigation by The Nerve in January 2026 identified at least 34 current and past UK state contracts across at least ten departments, worth at least £670 million, covering the Ministry of Defence, Highways England, DEFRA, the Home Office, the Government Digital Service, local councils including Coventry City Council, and multiple police forces.
The NHS FDP is the piece that has attracted the resistance. The contract terms, as characterised in Parliament by MPs citing the agreement, leave the NHS without ownership of the specially written software, without the intellectual property rights, and without the know-how transfer that would normally accompany a contract of this value. NHS England contests this framing and says Palantir has no rights in NHS data or in the commissioned canonical data model, while retaining IP only in the underlying Foundry product. Both statements can be simultaneously true; the contested question is what happens to operational capability if the contract ends, and the answer, acknowledged by the health minister at Westminster Hall, is that the NHS would need an alternative provider to be in place.
Amnesty International, Privacy International, the Good Law Project, Corporate Watch, and Medact have formed a No Palantir in the NHS coalition explicitly linking the FDP contract to Palantir’s work with ICE and the Israeli military. Approximately 200 NHS trusts had signed up to the FDP as of early 2026, but only around half were live and only a quarter reported benefits from using the system.
The UK case demonstrates two things the Canadian case cannot yet demonstrate. First, that a unitary state is not structurally better positioned to reverse a Palantir dependency than a federal one. The NHS is nominally a single institution, but 200 trusts making individual adoption decisions produces the same coordination problem as thirteen Canadian provinces making individual procurement decisions. Dispersion is a property of the delivery model, not only of the constitutional architecture. Second, that political resistance can become sustained and cross-party when the vendor’s external activities (ICE, Israel) are visible enough to generate public pressure. In the UK, the resistance is coming from Parliament, the health sector, human rights organisations, and medical professional bodies simultaneously. Whether the break clause is exercised in spring 2027 is the live question.
Canada: standing offer agreement, minimal political resistance
Canada is at Phase 3 on the staircase. The federal Supply Arrangement admitting Palantir to the procurement catalogue runs from January 2024 to July 2028. The Ontario Provincial Police contract is active. The Department of National Defence use is officially discontinued but was only disclosed to Parliament five years after signing. The lobbying pathway that established the federal relationship involved a former Ambassador to the United States who was found to have violated the Conflict of Interest Act but not the Lobbying Act for the same underlying conduct, and who now sits on the Prime Minister’s Council on Canada-US Relations.
The structural gap between the Canadian case and the UK case is that Canada has no Westminster Hall debate. No cross-party coalition has formed around the issue. No human rights organisation has tied the federal procurement to Palantir’s external activities. No provincial legislature has questioned the OPP arrangement at the level of sustained public pressure. The dependency is being built on schedule, and the window in which it could be interrupted before Phase 4 is the window that is open right now.
This is the context in which the Prime Minister’s “weakness” address has to be read. The speech act committed the government to structural correction. The procurement reality that would test the commitment is the Supply Arrangement extending to 2028, through which federal departments can continue to onboard the platform without triggering the competitive processes that generate parliamentary attention.
Germany: the partial reversal
Germany is the jurisdiction that demonstrates a Layer 5 sovereignty instrument actually working, at least partially.
The state of Hessen adopted Palantir’s Gotham platform, renamed “hessenDATA,” for police use beginning in 2017. Bavaria followed. In February 2023, the German Federal Constitutional Court ruled that the Hessen deployment’s legal basis was unconstitutional, on data protection and proportionality grounds. The ruling did not strike down Palantir as a vendor. It struck down the statutory authorisation under which police were using the platform, requiring Hessen to rewrite the enabling legislation before the deployment could continue.
This is a meaningful sovereignty outcome, and it is also a limited one. The court did not address the vendor relationship as such. It addressed the domestic legal architecture governing how the platform could be used. The deployment continued after Hessen amended the law. But the ruling establishes a precedent: a constitutional court in a G7 democracy has found that police use of a commercial data-fusion platform requires specific statutory grounding that ordinary administrative authorisation cannot supply. This is Layer 5 doing what Layer 5 is supposed to do: forcing the sovereignty question into legislative debate rather than leaving it to procurement paperwork.
No comparable Canadian, UK, or US ruling exists. The German case is the only one where the judicial branch has constrained the deployment pathway.
France: the case that declined
France is the jurisdiction that can be placed before Phase 1 on the staircase, because it reached the consideration stage and stepped back.
When the French Health Data Hub was being designed in 2019 and 2020, Palantir was reportedly among the candidates considered as a technical partner, in conjunction with Microsoft Azure as the underlying cloud infrastructure. The proposal generated sustained opposition from the CNIL (the French data protection authority), the Senate, and medical professional bodies on sovereignty and data protection grounds. The French government ultimately moved the Health Data Hub onto a domestic cloud infrastructure track, and Palantir was not contracted for the project.
The French case matters because it shows the dependency is not inevitable. A national government facing institutional pressure from its data protection authority, its legislature, and its professional bodies can decline the vendor at the consideration stage, before the Phase 1 crisis entry establishes a foothold. What this requires is a data protection authority with standing, legislative oversight that is actively engaged, and professional bodies willing to speak on the record. France had all three. The procurement decision followed the institutional architecture.
What the staircase says
The five-jurisdiction comparison makes three claims that no single-country case can make.
First, the dependency is a process, not an event. Each phase is distinct, each phase makes the next one easier, and each phase makes reversal more expensive. A country at Phase 3 is not facing the same problem as a country at Phase 4; it is facing an earlier version of the same problem, with more instruments still available.
Second, institutional architecture governs the outcome more than constitutional form. The UK is a unitary state and is facing coordination problems at the NHS trust level. Canada is a federation and is facing coordination problems at the provincial level. Germany’s reversal came through the judicial branch, not the federal executive. France’s decline came through the data protection authority working alongside the Senate. What distinguishes the jurisdictions that have pushed back from the jurisdictions that have not is whether the institutional instruments (courts, data protection authorities, parliamentary committees, professional bodies) are awake and engaged. Constitutional form is less determinative than institutional alertness.
Third, the vendor’s ideological disclosure on April 19, 2026 is a different kind of fact for each jurisdiction. For the United States, it is congruent with the procurement arc already complete. For the United Kingdom, it arrives in the middle of an active political debate and adds weight to the argument for exercising the break clause. For Canada, it is notice that any federal department procuring through the Supply Arrangement after this date is procuring from a vendor whose political commitments are declared rather than inferred. For Germany, it is a factor the Federal Constitutional Court may be asked to consider in any future challenge. For France, it is post-hoc vindication of the 2020 decision.
The 48-hour window between Saturday evening and Monday afternoon did not create the dependency. It made it visible. The Prime Minister of Canada named the weakness. The vendor identified, named into being, almost right on cue, exactly where the weakness exists. The UK Parliament named the exit option. The question that survives the weekend is which jurisdictions will use the instruments they still have.
The Canadian tell
One last observation, for the Canadian reader specifically. The Pentagon is testing redundancy against Palantir because it does not want its own strategic posture defined by a single vendor’s commercial and political trajectory. That is the most sophisticated customer in the world, with the most advanced dependency, openly hedging.
A Canadian federal department procuring under the Supply Arrangement between now and July 2028 is not hedging. It is deepening. That is the operational gap the Prime Minister’s address identified without naming. The staircase shows where the gap leads.
Sources and dated claims: US Army press release (31 July 2025); Wired, CNBC, Breaking Defense, Defense One, Washington Technology, Defense Scoop (multiple reports 2025-2026) on the Army Enterprise Agreement and DHS BPA; SpaceNews and Tom’s Hardware on Maven contract ceiling and program of record designation; Palantir Q4 2025 earnings release; CBS News, CNBC, OpenSecrets, Wikipedia on Vance-Thiel-Mithril relationship and Protect Ohio Values super PAC; The Register (20 April 2026) on Westminster Hall debate; The Lowdown, NHS England FDP contract explainer, Pharmaphorum, Amnesty International UK on the NHS FDP contract; The Nerve (January 2026) investigation of UK Palantir contracts; German Federal Constitutional Court ruling (February 2023) on Hessen police use; public record on the French Health Data Hub decision. Palantir Technologies social media post (19 April 2026) summarising excerpts from The Technological Republic: Hard Power, Soft Belief, and the Future of the West by Alexander C. Karp and Nicholas W. Zamiska.