The Manifesto They Meant to Say Out Loud
What Palantir's 22-point declaration reveals about the cost of dependency — and why democratic nations are quietly building the exit
John Pope · Founder, midagent AI · Ottawa, Canada · April 2026
I have written in this space about Alex Karp with genuine praise. His argument about neurodivergence and the future of work is one of the more honest and important things a technology executive has said in public in years. I meant it when I said so. I am not one of those writers who discovers that someone was always wrong the moment they say something inconvenient.
So let me be equally honest about this.
The 22-point manifesto that Palantir posted to X last weekend — a distillation of Karp's book The Technological Republic, co-authored with Nicholas Zamiska — is not a bold vision. It is not a provocation designed to generate useful debate. It is a political document, written with deliberate architectural precision, that announces something its authors have apparently decided no longer needs to be concealed.
And what it reveals, for any government currently running Palantir software on its most sensitive infrastructure, is a problem that has nothing to do with the quality of the product.
What the Manifesto Actually Says
The 22 points cover considerable ground. Some are anodyne. Some are, in isolation, defensible. That is not an accident — the moderate-sounding points are the scaffolding that makes the emotionally-charged, load-bearing ones easier for the audience to accept.
The load-bearing ones include: that Silicon Valley has an "affirmative obligation" to participate in the defence of the nation; that hard power in this century will be built on software; that national service should be a universal duty; that the postwar constraints on German and Japanese military capacity "must be undone"; that the West must resist what it calls a "vacant and hollow pluralism"; and that some cultures have produced "vital advances" that others have not.
Mark Coeckelbergh, a philosopher of technology at the University of Vienna, called it "technofascism." Yanis Varoufakis described it as a willingness to add "the AI-driven threat to humanity's existence" to the existing arsenal of catastrophe. The University of Ottawa's academic community joined a growing chorus of scholars and civil society organisations who found in the manifesto something that went far beyond a corporate mission statement.
I do not think Karp is a fascist. I do think he is a man who has become very comfortable saying, at significant volume, things that would previously have been said only in rooms with thick walls and no windows. And I think the reason for that comfort is not hard to identify.
The Permission Architecture
The Trump administration did not create Palantir's ideology. But it created the permission architecture for Palantir to declare it openly.
When the Oval Office frames geopolitics as civilisational conflict, when the Department of Defence becomes, in all but name, a Department of War, when the language of American dominance shifts from diplomatic aspiration to explicit policy — the companies that have built their entire business model on that dominance no longer need to perform ambivalence. The quiet part is no longer quiet because the people at the top of the American government have decided that loudness is, itself, a form of power.
Karp is not echoing the manifesto into a void. He is echoing sentiment from the highest levels of an administration that has deployed tariffs against allies, acknowledged the use of intelligence resources in bilateral trade negotiations with Canada, and framed democratic middle powers as raw material suppliers rather than sovereign partners. The manifesto is not a departure from that worldview. It is its software expression.
This matters for a specific and practical reason. Palantir is not merely a company with a controversial CEO. It is the company whose software runs inside the NHS. Inside the UK Ministry of Defence. Inside Canadian federal departments. Inside the intelligence infrastructure of five-eyes nations. It holds contracts worth hundreds of millions of pounds across more than thirty UK government bodies alone. It is, in a meaningful sense, embedded in the nervous system of democratic governance in the English-speaking world.
And its outspoken leader has just told you, in 22 points, what the company advocates for and believes with conviction.
The whistleblower account below describes the methods and culture inside Palantir.
The Lock-In Is the Problem
I want to be precise about why the manifesto matters beyond the ideological content.
If Palantir were a company with controversial politics and no particular structural leverage over its customers, the appropriate response would be to note the controversy and continue the procurement conversation on its merits. Companies have complicated founders. Life is complicated.
But Palantir is not that kind of company. Its contracts are structured, by design, to make departure expensive and technically prohibitive. The NHS contract — the £330 million Federated Data Platform agreement now being nervously re-evaluated by UK ministers — specifies that all intellectual property for specially written software belongs to Palantir. Every line of code written to connect NHS trust systems to the platform belongs to Palantir. Every custom integration, every data mapping, every analytical schema built over years of deployment — Palantir's. When a government customer leaves, they leave without their institutional memory. The switching cost is not financial. It is epistemological: you lose the intelligence your own data has generated, because the platform that processed it owns the conclusions.
This is not a design flaw. It is a design choice. And it is a design choice that makes the manifesto's ideology not merely a reputational concern but a structural one.
A democratic government that has embedded Palantir in its critical health, defence, and intelligence infrastructure is not just buying a technology product. It is creating a dependency on a company that has now stated, openly, that it sees its role not as serving its customers' interests but as advancing a specific civilisational agenda — one shaped by and aligned with a particular American political moment. That agenda may shift. The lock-in will not.
Labour MP Dawn Butler put it plainly in the Westminster Hall debate last week: "We must ask ourselves whether Palantir allows us to have AI and data sovereignty. I think the answer to that will be no."
She is right. And the answer was always going to be no — not because of the manifesto, but because of the architecture beneath it.
The Procurement Reckoning That Is Already Happening
The UK government is, as of this writing, quietly exploring the technical process of removing Palantir from NHS systems ahead of the formal break clause review scheduled for early 2027. The Health Minister has told Parliament that if other providers "can do the job better," that will be "reflected upon." UK Science Minister Lord Vallance has testified before the Commons Science Committee with a message of notable clarity: the future is "putting British companies there and procuring innovation here."
This is not isolated. Switzerland's army has formally recommended its government consider alternatives to Palantir, citing fears that sensitive military data could be accessible to American intelligence agencies. The British Medical Association — representing more than two hundred thousand physicians — has urged its members to limit engagement with the NHS Federated Data Platform. Forty-seven thousand patients have signed petitions opposing the contract. Protests have been staged outside Palantir offices in New York, Aventura, and across Europe.
The manifesto did not create this movement. But it has crystallised it. It has converted a diffuse unease — about a US company with intelligence agency roots handling the most sensitive data of allied democratic governments — into a specific, named, documented concern. You can now point to 22 numbered propositions and ask, plainly: is this the values framework of the company you want embedded in your national health service?
The reckoning was coming. The manifesto accelerated it.
What Democratic Nations Actually Need
This is where I want to be careful, because there is a version of this argument that is merely reactive — that says "not Palantir" without being able to say "what instead."
The need that Palantir addresses is genuine. Democratic governments need sophisticated data analytics platforms. They need the ability to federate data across agencies and jurisdictions, to generate policy-grade intelligence from complex and sensitive datasets, to make AI-powered decisions that are auditable, explainable, and legally defensible. These are not trivial requirements. The fact that Palantir built a product that meets them — and that no obvious democratic alternative yet exists at comparable scale — is why governments signed the contracts they signed, even when the politics of the company were already uncomfortable.
But the requirements have changed. The manifesto has made visible something that was always structurally true: a platform governed by US law, built by a company with CIA origins and explicit ideological commitments to American hard power, cannot serve as neutral infrastructure for allied sovereign governments. It was never truly neutral. The question was always whether the dependency was worth the political cost. The manifesto has changed the calculus on the political cost side of that equation, dramatically and permanently.
What is needed, and what forward-looking governments are beginning to articulate, is an alternative that does not simply replace one form of dependency with another. Not European dependency instead of American. Not Canadian dependency instead of British. Genuine architectural sovereignty: a platform built on open standards, operating under domestic legal jurisdiction, structured so that the data it processes and the intelligence it generates belong permanently to the government that generated them — not to the platform that processed them.
The architecture requirements are specific. Protected data must stay under domestic legal jurisdiction, with no exposure to foreign compulsion orders. The reasoning engine must operate transparently, with auditable outputs and documented data lineage. The platform must be built on open standards — not proprietary schemas that create lock-in at the data layer. The governance structure must make acquisition by a dominant foreign platform structurally impossible, not merely contractually discouraged. And the intellectual property generated must belong to the customer, not the vendor.
These are not aspirational requirements. They are the minimum conditions for a sovereign government to trust that its most sensitive analytical infrastructure is genuinely under its control.
The Alternative Is Being Built
I will not pretend that a fully operational alternative to Palantir Foundry exists today at the scale that democratic governments require. It does not. The honest accounting of where we are is this: the procurement vacancy is live and widening, the political pressure to fill it is intensifying, and the architectural specifications for what a genuine democratic alternative would require are well understood. What is happening now — in Canada, in the UK, in Germany, in the allied nations that have signed digital sovereignty cooperation agreements over the past eighteen months — is the translation of those specifications into real platforms.
Canada is further along this path than the public conversation acknowledges. The foundational architecture for a Protected B-certified, open-standard, CLOUD Act-immune analytics platform — one built on Canadian-owned infrastructure, governed by Canadian law, and structured so that no foreign authority can compel access to the data it processes — is not a theoretical proposal. It is in development. It is built on the same open standards that make Palantir's proprietary alternatives increasingly anomalous. Its governance architecture embeds the public-interest mandate in the legal structure of the institution, not in the discretion of a founder or a board.
It embodies something that Palantir, in 22 numbered points, has confirmed it does not: a commitment to the values of the democratic governments it would serve, rather than the ideological agenda of the company that built it.
That is not a small distinction. In the current environment, it may be the only distinction that matters.
The Quiet Part, Now Loud
I want to end where I began — with Karp himself, because I think intellectual honesty requires it.
Alex Karp is a genuinely brilliant and genuinely complex figure. His neurodivergent advocacy is real and his analysis of the AI labour market is sharper than most. I do not think he woke up one morning and decided to become a villain. I think he believes what he wrote. I think he believes, with real conviction, that democratic civilisation requires hard power, that software is the instrument of that power, and that the time for polite ambiguity about this has passed.
I disagree with him. Completely and without reservation.
Not because hard power is irrelevant — it is not. Not because the geopolitical threats to democratic nations are imaginary — they are not. But because the architecture of a surveillance platform embedded in the health, defence, and intelligence systems of democratic governments cannot be the personal ideological property of any single company, any single founder, or any single political moment. Democratic infrastructure must be accountable to democratic institutions. That is not a soft or naive position. It is the foundational requirement of democratic governance, and no quantity of civilisational urgency excuses the abandonment of it.
The manifesto was, in a sense, a gift to those of us who have been making this argument. Palantir has now said, plainly and publicly, that its platform is not neutral infrastructure in service of its customers' sovereign interests. It is the technological expression of a specific political vision — one that happens to align very closely with the current occupants of the White House and the Pentagon.
Every government that is currently running Palantir software on its most sensitive systems now has a decision to make. The decision is not primarily technical. It is political, in the deepest sense of that word: a decision about whose values are encoded in the infrastructure of democratic governance.
The alternative is being built. The specification is clear. The window is open.
The question is whether the governments that need it most will move before it closes.
John Pope is the Founder and Lead Strategist of midagent AI, Ottawa, Canada. The Innovation Nation Letters is published at midagent.ca/blog · hello@midagent.ca
Comments