A forensic reading of the public record, and what it means when we line it up against the layers.
An unusual disclosure, on an unusual day
Two things happened on Sunday, April 19, 2026, that should be read together.
In a pre-recorded address to Canadians released that morning, Prime Minister Mark Carney told the country that the Canada-U.S. relationship, for generations a pillar of Canadian prosperity and security, has become a weakness. He said his government is making structural changes to “make us stronger at home and less reliant on the United States.” He invoked a statue of Major-General Sir Isaac Brock, who died defending Niagara from the American invasion of 1812, as the visual anchor of the address. He told Canadians he would “never sugarcoat our challenges,” and he named the challenge: the bond with America, the thing that used to be the strength, is now the thing to correct.
The same day, Palantir Technologies posted a summary of its CEO’s book manifesto, The Technological Republic: Hard Power, Soft Belief, and the Future of the West, by Alexander C. Karp and Nicholas W. Zamiska, on social media, framed as a set of 22 numbered theses. The company prefaced it with “because we get asked a lot.”
The document is worth reading carefully, because it is not the usual corporate communication. It is a political platform published under a vendor’s brand. The theses include the assertion that “Silicon Valley owes a moral debt to the country that made its rise possible” and has “an affirmative obligation to participate in the defense of the nation”; that “the question is not whether A.I. weapons will be built; it is who will build them and for what purpose”; that “the postwar neutering of Germany and Japan must be undone”; that “some cultures have produced vital advances; others remain dysfunctional and regressive”; and that the West must “resist the shallow temptation of a vacant and hollow pluralism.”
Most vendors do not publish foreign-policy doctrines. Most vendors do not argue against pluralism, call for the re-militarization of Germany and Japan, advocate for universal conscription in the United States, or position themselves as the software layer of a new deterrence regime replacing the atomic one. Palantir has now done all of those things, in writing, under its own name, on the record. This is a positioning decision, not a disclosure accident. Palantir has elected to be read as a political actor rather than as neutral infrastructure, a choice to trade the strategic ambiguity vendors normally prefer for explicit political legibility, on the record, under its corporate brand. For every allied state that procures the platform, the vendor relationship is now configured differently than it was last week. The question is no longer what Palantir believes. It is why the company has decided the moment has come to say so plainly.
These two Sunday-morning disclosures define the argument of this piece. The Prime Minister of Canada has stated, in a direct address to the nation, that reliance on the United States is a weakness requiring correction. A U.S. defence-software vendor holding pre-approved status on the Canadian federal procurement list through 2028, and with dated, disclosed contracts at the Department of National Defence and the Ontario Provincial Police, has stated, in a political platform published under its corporate brand, that it is the software layer of a U.S. national-security project built on hard power, cultural hierarchy, and the rejection of pluralism.
This matters for Canadian procurement because it collapses a distinction that normally protects buyers. A software license is usually understood as morally neutral infrastructure, like electricity, or plumbing. The vendor sells the tool; the buyer decides the use. What Palantir has published today is an explicit argument that its products are not neutral infrastructure. They are instruments of a specific political project: hard power over soft, cultural hierarchy over pluralism, military readiness over deliberation, and the rejection of democratic pacing as a luxury the West can no longer afford.
Canadian institutions that have procured Palantir software, and those that can now procure it through the 2024–2028 Supply Arrangement described below, are not merely licensing a data-fusion platform. They are contracting with a company that has, as of today, formally declared what it believes the platform is for. By the Prime Minister’s own framing, delivered the same day, that is the definition of the weakness that requires correction. That is the context in which the disclosure record needs to be read.
The disclosure record, assembled
There is no Canadian registry of Palantir deployments. What exists is a set of disclosures extracted one at a time, through written questions on the Order Paper, access-to-information requests, investigative journalism (mostly by The Logic), and the occasional ethics commissioner report. Assembled together, they describe a procurement pathway that operated largely outside parliamentary and public oversight for nearly a decade.
Federal: National Defence
The federal anchor contract is the Department of National Defence’s $14.4-million agreement with Palantir Canada, signed in March 2020 for the Gotham data processing and analytics platform. The government did not disclose this contract at the time of signing. It surfaced only when an MP registered a written question in the House of Commons, and the department tabled its answer in Parliament in September 2025, more than five years later. National Defence says it is no longer using the software. It also used Palantir’s Foundry platform intermittently between May 2022 and March 2024 for “an evaluation activity to assess capabilities related to data processing.”
That March 2020 contract was preceded by a $997,434 one-year deal signed in March 2019 for software used by the Canadian Special Operations Forces Command (CANSOFCOM), the units that respond to hostage situations and chemical, biological, radiological, and nuclear incidents. This was the first Palantir procurement the federal government ever disclosed. It was non-competitive. Other vendors did not bid.
The March 2019 contract is itself worth examining. In 2017, both the Department of National Defence and Public Safety Canada told the CBC they had never bought or used Palantir’s technology, despite the CBC identifying two former Public Safety employees who listed using Palantir software as part of their duties on LinkedIn. When asked about this in 2019, Public Safety stated Palantir was “not used at Public Safety, nor is it a part of our approved software list.” The RCMP and the Communications Security Establishment declined to answer whether they use Palantir’s products. Those positions have never been updated on the public record.
Federal: the pre-approved vendor list
In 2019, Public Services and Procurement Canada added Palantir to its list of pre-qualified artificial intelligence vendors. In January 2024, Palantir was awarded a 53-month Supply Arrangement running from January 2024 to July 2028, making its licenses widely available to federal departments, agencies, Crown corporations, and other authorized clients through the Government of Canada’s Software Licensing Supply Arrangement (SLSA). Carahsoft was named the sole distributor.
What a Supply Arrangement Actually Is:
Some terminology is worth unpacking here, because the procurement vehicle is doing most of the work that matters in the Palantir case, and most readers will not know what it means when a vendor is described as being “on the SLSA.”
Canadian federal software procurement runs through Public Services and Procurement Canada (PSPC), the central purchasing department that handles buying for the rest of the federal government. PSPC uses several pre-qualification vehicles to speed up procurement. The most common ones are Standing Offers and Supply Arrangements. Both work on the same basic principle: instead of requiring every department to run a full competitive procurement every time it wants to buy something, PSPC runs one big qualification process up front, and any department that wants the qualified product afterwards can use a simplified process to buy it.
The Software Licensing Supply Arrangement (SLSA) is the specific vehicle that covers commercial software licenses for the federal government. It operates in two phases. In Phase 1, a vendor responds to a public Request for Supply Arrangements, gets evaluated against PSPC’s criteria, agrees to a price ceiling and to the government’s standard terms and conditions, and, if it qualifies, gets its software listed in the SLSA catalogue. That catalogue is only accessible to federal government employees. In Phase 2, any federal department that wants to buy the software uses a simplified purchasing process to do so. Depending on the circumstances, this can be a scaled-down Request for Quotations, a simplified bid solicitation, or, if the mandatory sole-source justifications are met, a direct contract with no competitive process at all.
The accountability consequence: once a vendor is on the SLSA, the decision about whether to allow that vendor access to Canadian federal departments has already been made. The remaining decisions (which departments actually buy, how much, for what purpose) happen under a simplified regime that does not typically attract parliamentary attention, media coverage, or public consultation. The heavy procurement lifting is front-loaded into Phase 1. Individual deployments in Phase 2 often proceed through what amounts to administrative paperwork.
Carahsoft, a U.S. government-IT distributor headquartered in Reston, Virginia, holds the SLSA contract through which Palantir software is made available to the Canadian federal government. In January 2024, Palantir and Carahsoft announced that Carahsoft had been authorized as the sole distributor of the Palantir Platform under the SLSA. This is the 53-month arrangement running from January 23, 2024 to July 4, 2028 that appears in the CanadaBuys contract history. What that means, in plain terms: any Canadian federal department, agency, or Crown corporation that wants to procure Palantir’s data-fusion platform between now and July 2028 can do so through an existing pre-qualified vehicle, without triggering a new competitive procurement, and without the kind of disclosure obligations attached to a standalone contract. The decision to admit Palantir to the federal software ecosystem was made in 2024, through a procurement vehicle most Canadians have never heard of, and its consequences will propagate through whatever Phase 2 purchases occur before 2028.
Supply Arrangements are, in principle, a legitimate tool. They exist for good reasons. Government procurement is slow, and bulk pre-qualification speeds up legitimate software purchasing for everything from Adobe Acrobat to enterprise databases. The problem is not the tool. The problem is that the tool was designed for commercial software with uncontroversial political implications (word processors, antivirus suites, project management platforms) and is now being used to onboard a vendor whose CEO has, as of April 19, 2026, published a 22-point political platform as corporate communication. The vehicle was not built to ask whether a vendor’s declared ideology is compatible with Canadian public values. It assumes that question does not need to be asked. That assumption was defensible in 2013 when the SLSA’s current form was established. It is no longer defensible now.
This is the piece of the map that changes the character of the question. A single $14-million contract is a procurement event. A pre-approved standing vendor agreement extending to 2028 is infrastructure. It means any federal department wanting to procure Palantir can do so through an existing vehicle, without the competitive process, the parliamentary scrutiny, or the political friction that a standalone contract would attract. The surface area of potential deployment is now the entire federal government, and the disclosure obligations attached to individual uses are thin.
Provincial: Ontario
The Ontario Provincial Police (OPP) is the clearest provincial case. In October 2022, The Logic reported that the OPP was using Palantir’s Gotham platform for crime analysis. The force confirmed use but would not describe how it was being deployed. The original OPP contracts were marked “not for public disclosure” and were covered by a section of Ontario’s Financial Administration Act that allows the provincial solicitor general to authorize law-enforcement spending without further examination, effectively a statutory exemption from the normal accountability regime.
Former Ontario Community Safety Minister Yasir Naqvi later acknowledged the OPP had not consulted him before signing contracts with Palantir.
After public attention faded, the Ontario government issued a Request for Qualifications in September 2022 for a data analytics platform that could potentially replace Palantir. Nine vendors qualified. One year later, in October 2023, Palantir was awarded a $36.6-million contract. Media coverage of the award was sparse.
Provincial and municipal: mostly dark
Beyond the OPP, the provincial and municipal record is largely not public. Journalists and civil society organizations have referenced Palantir’s “quiet integration” at federal, provincial, and municipal levels, but specific contracts at other police services, health authorities, or city governments have not been systematically disclosed. During the COVID-19 pandemic, David MacNaughton (then Palantir Canada’s president) stated publicly that the company was working with the federal government and several provinces on pandemic response. Which provinces, and on what, was never detailed on the public record.
The lobbying record
In August 2019, David MacNaughton was appointed president of Palantir Canada. He had just stepped down as Canada’s Ambassador to the United States, where he had served as Prime Minister Trudeau’s point man in Washington for three and a half years.
Between February and April 2020, MacNaughton arranged multiple meetings with senior federal officials, including Deputy Prime Minister Chrystia Freeland, Innovation Minister Navdeep Bains, and Chief of Defence Staff General Jonathan Vance, to offer Palantir’s pro bono assistance with Canada’s COVID-19 response. In September 2020, federal Ethics Commissioner Mario Dion ruled that MacNaughton had broken the Conflict of Interest Act. The commissioner’s order detailed 17 communications with nine public office holders and prohibited all nine from having official dealings with MacNaughton for one year. MacNaughton consented to the order and acknowledged the contacts contravened the Act.
In March 2021, the federal Lobbying Commissioner issued a separate finding: MacNaughton had not broken the Lobbying Act’s five-year post-employment restriction. The reason was procedural. Most of his contacts with senior officials concerned possible pro-bono contracts, which are not covered by the lobbying restriction. On the communications that were covered, he spent just over two and a half hours, below the 20 percent “significant part of work” threshold that triggers the prohibition. The commissioner renewed her call for Parliament to close the loophole.
MacNaughton was found to have violated one federal accountability statute and not another, for the same underlying conduct. His presidency of Palantir Canada is now a past role. His public biography describes him as a Strategic Advisor in the Office of the CEO at CIBC, but he sits as a member of the Prime Minister’s Council on Canada-U.S. Relations, confirmed by the Prime Minister’s Office on March 19, 2025. Palantir continues to win contracts.
Mapping to the sovereignty framework
This is where the map stops being a list of procurement events and starts describing a systemic failure. The dependency framework we’ve been developing, the four triggers, the three cost layers, the five-layer sovereignty gap model, reads the Palantir record in a specific way.
The four triggers, instantiated
The framework identifies four conditions that convert vendor relationships into sovereignty exposures: opacity, irreversibility, foreign jurisdictional reach, and institutional capture. The Palantir record satisfies all four.
Opacity. The five-year gap between the 2020 DND contract signing and its 2025 disclosure is the canonical case. The OPP’s statutory exemption under the Financial Administration Act is opacity codified into law. The “not for public disclosure” designation on the original OPP contracts is opacity as procurement default. A dependency the public cannot see is a dependency the public cannot consent to, cost, or constrain.
Irreversibility. Data fusion platforms like Gotham and Foundry are not substitutable in the way that, for example, an email system is. They ingest heterogeneous datasets, build integration logic, train operators, and embed themselves in analytical workflows. Replacing them requires rebuilding the fusion layer, retraining staff, and migrating the institutional memory encoded in the platform’s ontology. The Ontario RFQ that ran for a year before re-awarding to Palantir is evidence of this: the competitive process notionally existed, but the switching cost favored the incumbent. This is the attrition trap, expressed in procurement.
Foreign jurisdictional reach. Palantir is a U.S. company. Its largest customer is the U.S. federal government. It is reportedly being used by the Trump administration to manage and share data within and between departments. Its products have been used by U.S. Immigration and Customs Enforcement for raids and family separations. None of this is a comment on the specific Canadian deployments. It is a comment on the jurisdictional regime governing the vendor. Under the CLOUD Act and related U.S. statutory authorities, data processed on U.S.-company infrastructure is reachable by U.S. agencies in ways that do not require Canadian legal process. A Canadian sovereign dependency that routes through a U.S. vendor is, by construction, a Canadian sovereign dependency that routes through U.S. legal authority. This is the Layer 5 provincial sovereignty gap made concrete.
Institutional capture. The MacNaughton pathway is textbook. Ambassador to the U.S., steps down, takes the presidency of the U.S. vendor’s Canadian subsidiary within weeks, secures meetings with the senior officials he worked alongside months earlier, offers pro-bono services during a crisis when normal procurement discipline is relaxed, and is found to have violated the Conflict of Interest Act but not the Lobbying Act because the two statutes are inconsistent. The gap between the two rulings is the loophole. The loophole has not been closed.
MacNaughton’s presidency of Palantir Canada is now a past role; his public biography describes him as a Strategic Advisor in the Office of the CEO at CIBC. More consequentially for the sovereignty argument, he is a current member of the Prime Minister’s Council on Canada-U.S. Relations, the advisory body the Prime Minister’s Office confirmed on March 19, 2025, which counsels the Prime Minister and Cabinet on the Canada-U.S. trade and security relationship, including on tariffs, counter-tariffs, and the strategic posture toward the United States. The Council role places a former president of the Canadian subsidiary of a U.S. defence-software vendor inside the room where Canada’s U.S.-facing strategy is being advised. Whether or not there is any current commercial relationship, the institutional proximity (former vendor president, current advisor to the Prime Minister on U.S. relations) is the kind of configuration a functioning sovereignty regime would surface and disclose on the record, not leave to inference.
A fifth condition the original framework did not name: declared vendor ideology. We currently refer to this as “change in temperature” but today’s Palantir posting forces this further and more explicitly into the analysis. The framework assumed vendor opacity, that we would have to infer the political orientation of a platform provider from its customer list, its executive speeches, and its products’ downstream uses. Palantir has removed that analytical step. It has published, as corporate communication, a 22-point political platform advocating hard power, cultural hierarchy, and the rejection of pluralism. For a Canadian institution, the question whether the vendor might have commitments incompatible with Canadian public values has an answer it offered up itself. The vendor has stated those commitments. A procurement decision made after April 19, 2026, is a procurement decision made with notice.
The three cost layers
The framework distinguishes direct procurement costs (what the contract says), institutional costs (capacity displaced), and sovereignty costs (decisions the vendor now meaningfully constrains).
The $14.4M DND contract and the $36.6M OPP contract are direct costs, and they are small. They are small enough that they fall below the threshold at which most parliamentary committees or legislative finance offices will examine them in detail. This is part of the design. The institutional cost is larger: every year the Gotham platform is in place at DND is a year that internal Canadian data-fusion capacity is not being built. The sovereignty cost is the hardest to quantify but the most consequential. Canadian defence and policing decisions are now mediated by analytical infrastructure that a foreign vendor controls, on pricing and availability terms that are not sovereign to Canada.
The five-layer gap
The sovereignty gap model locates the failure at each layer of the stack. Palantir hits at least four:
- Layer 1 (hardware and compute): The platforms run on cloud infrastructure that is overwhelmingly U.S.-hyperscaler. DND’s claim that Gotham was used only on “closed, classified” networks within secure facilities is plausible for that specific deployment; it is not generalizable across the Supply Arrangement.
- Layer 2 (model and platform): Palantir is the platform. The ontology, the integration patterns, the analytical primitives: all proprietary, all foreign.
- Layer 3 (data): Whatever data Canadian institutions load into Gotham or Foundry is, at a minimum, processed through software whose behavior is not auditable by Canadian authorities in the way domestic software would be.
- Layer 5 (institutional and procurement governance): This is where disclosure limitations, the statutory exemptions, and the MacNaughton pathway live. The existing institutional mechanisms — competitive procurement, lobbying restrictions, parliamentary oversight, were not sufficient to govern the relationship. The CBA-driven protections emerging in parts of the federal public service, which we’ve been tracking, do not yet extend to platform procurement decisions of this kind.
A federalism note, and an operational weakness the framework has to name: the discussion above slides between “Canada,” “the federal government,” and “Canadian institutions” in ways that understate how much of the Palantir exposure sits outside federal jurisdiction. Policing is provincial. Health is provincial. Conflict of interest and lobbying rules are distinct statutes at the federal level and in each province, governed by separate commissioners, separate thresholds, and separate enforcement regimes. There is no mechanism by which Ottawa can harmonize lobbying or conflict-of-interest rules across the federation, and no prospect of one. Constitutional division of powers forecloses it.
This matters operationally, not just procedurally. The federal government can, in principle, act on its own exposure quickly: it can legislate to cancel the Supply Arrangement, require disclosure of existing deployments, direct departments off the platform, and close the Lobbying Act loophole the Commissioner has twice asked Parliament to close. Those are all federal decisions, and the federal government can and probably should make them. The Ontario Provincial Police contract is not a federal decision. The statutory exemption in Ontario’s Financial Administration Act that shielded the original OPP contracts from disclosure is an Ontario statute. Any provincial health authority that has procured or may procure the platform sits under provincial accountability law. A federal response, even a decisive one, leaves the provincial surface area intact.
This is an operational weakness the sovereignty framework has to name explicitly. The four triggers and five-layer model were developed against the federal case, where at least the accountability instruments, however inconsistent, are held by a single Parliament. At the provincial level, there are thirteen separate accountability regimes, no inter-provincial registry, no common disclosure standard, and no mechanism to compel one. A vendor that wins on the federal SLSA has, in effect, acquired a set of thirteen parallel provincial sales opportunities, each governed by rules the federal government cannot reach. The Prime Minister’s framing, that reliance on the United States is a weakness to correct, is a federal speech act. The correction, in any domain where the dependency has already crossed into provincial jurisdiction, requires thirteen separate political decisions that no one has the authority to coordinate.
This is jurisdictional fragmentation as a sovereignty condition in its own right, distinct from the four vendor-relationship triggers and from the five-layer stack gap. It describes the buyer’s own governance architecture, and in the Canadian case it is the condition that limits what any federal action, however decisive, can accomplish. The framework has to carry it as a named variable, not as background. This is nothing new in Canada, but it is an additional layer of complexity when dealing with technology governance.
The attrition trap, in real time
The framework’s attrition trap argument is that sovereignty erodes faster than institutional defenses can be negotiated. The Palantir record demonstrates the trap in operation. The Supply Arrangement runs to July 2028. The federal IT collective agreement expired in December 2025. CBA negotiations covering AI deployment clauses are under way but incomplete. The window in which a pre-approved vendor list can expand, departments can adopt through light-touch procurement, and the platform can embed itself in workflows. That window is open right now, and it is open because the protective institutional mechanisms are still being built.
This is the Ontario March 2026 medical records case in a different domain: the announcement with no timeline and no funding, the voluntary adoption framing, the dependency embedding while the governance is still being written.
What a Sovereign Exposure Registry would show
The Registry proposal we’ve recommended exists in part because the current disclosure regime cannot surface a dependency map like this one in real time. The Palantir record had to be reconstructed piece by piece, across five years of journalism and two parliamentary questions, before the shape became visible.
A functioning Registry would have, at minimum:
- A standing requirement that every federal department, agency, and Crown corporation disclose all contracts with foreign-headquartered AI and data-platform vendors, including those procured through Supply Arrangements, at the point of signing rather than on response to a written question.
- Mandatory disclosure of the jurisdictional regime governing each contract: which foreign laws can reach the data, under what process, with what notification to Canada.
- Provincial reporting obligations matching the federal standard, with specific coverage for policing and health deployments where the Ontario Financial Administration Act precedent currently permits opacity.
- A harmonization of the Conflict of Interest Act and Lobbying Act thresholds for former designated public office holders, closing the loophole the Lobbying Commissioner has twice asked Parliament to close.
- A Layer 2 audit requirement, independent Canadian assessment of the platform itself, not just the contract.
None of this exists. The pre-approved vendor list does. The Supply Arrangement does. The contracts do. The lobbying pathway does.
The Record Poses Questions
Canada has not debated whether to build sovereign data-fusion capacity, whether to accept a foreign vendor in that role, or what price in jurisdictional exposure is acceptable in exchange for what operational capability. Those debates did not happen because the underlying procurement decisions were not visible when they were being made. They became visible afterwards, in fragments, through the work of a small number of journalists and one persistent opposition MP.
The Palantir case is not unique. It is the most documented example of a pattern: a Canadian sovereign function (defence, policing, health data, potentially more) is quietly routed through a U.S. platform, under a procurement regime that does not require disclosure at the point of decision, governed by accountability statutes that are internally inconsistent, and embedded deeply enough that reversal is expensive by the time the public learns.
The sovereignty framework names this as a system, not a series of incidents. The disclosure record, read through the framework, is the evidence the system exists.
The Prime Minister, in his Sunday address, asked Canadians to accept that the former strength has become the weakness. The test of that framing is whether the government acts on it in the places where the weakness is already embedded: not in speeches and in trade diversification alone, but in procurement, in disclosure, in the Supply Arrangements that extend past the next election, in the Councils that advise the Prime Minister, and in the vendor ideologies that now come declared rather than inferred. The Palantir case is the first place to look, because it is the place where the system is already built.
Sources: The Logic (multiple reports 2019–2025), CBC News, Global News, the Office of the Commissioner of Lobbying of Canada (Investigation Report: David MacNaughton, President of Palantir Canada, 2021), the Office of the Conflict of Interest and Ethics Commissioner (Order, September 2020), CanadaBuys contract history (Palantir Technologies Canada Inc.), House of Commons written questions (Q-8, 45th Parliament, first session), Office of the Prime Minister of Canada (readout: Prime Minister Carney meets with Prime Minister’s Council on Canada-U.S. Relations, March 19, 2025), BNN Bloomberg (exclusive: Prime Minister Carney’s pre-recorded address to Canadians, April 19, 2026), Palantir Technologies social media post summarizing excerpts from The Technological Republic: Hard Power, Soft Belief, and the Future of the West by Alexander C. Karp and Nicholas W. Zamiska (April 19, 2026), Ricochet Media, Open Canada, Globe and Mail (May 2021), Carahsoft press release (January 2024).