Palantir released a manifesto on AI sovereignty and attacked the token-based AI monetization model

On June 30, 2026, Palantir published a nine-point manifesto on 'AI sovereignty' on the X platform and declared war on the dominant business model of the AI industry — monetization through token volume. The document was titled 'thoughts on the importance of AI sovereignty'.

Nine principles against token economics

The manifesto calls on organizations to stop relying on third-party cloud AI models and regain control over data, algorithms, and infrastructure.

Key positions:

• Organizations must accumulate and protect their own data without sharing it with third-party providers
• Model weights must belong to the organization itself — not rented through a subscription
• The main target of criticism is 'token-maxing': the practice of maximizing AI provider revenue through increased request volume and response length
• Sovereign AI implies deployment on the client's own infrastructure without dependence on external APIs

The term 'token-maxing' is the central concept of the document. Palantir points to a systemic conflict of interest: a cloud AI provider earns more when the model generates longer responses and when the API is called more frequently. This incentive structurally contradicts the interests of the client who wants accurate, short, and cheap answers. The more 'verbose' the AI is, the more profitable for the provider, but the more expensive for the organization.

Target audience of the manifesto

The target audience is clear: government agencies, defense departments, and large corporations — Palantir's traditional client base. Since 2003, the company has specialized in supplying analytical platforms to the governments of the United States, the United Kingdom, and NATO partners.

For these clients, data sovereignty is not rhetoric but an operational necessity. Processing intelligence data, industrial secrets, or citizens' personal data through a third-party cloud provider's infrastructure creates legal and political risks incompatible with regulatory requirements.

Palantir's market logic, however, is transparent: the company is a direct beneficiary of the transition to 'sovereign AI'. The publication TNW characterizes the manifesto without mincing words — 'a proposal disguised as a principle'.

'This is a proposal disguised as a principle' — TNW on

Palantir's manifesto.

Who comes under attack

The manifesto implicitly attacks the business models of leading AI providers — OpenAI, Anthropic, Google, and Microsoft. Their revenue directly depends on token volume: the more requests the system processes, the higher the company's income. Palantir argues that this mechanism creates incentives that contradict the interests of corporate and state clients.

Palantir offers an alternative: own infrastructure, own model weights, own data. Initial investments are higher, but the organization gains independence from a specific vendor's conditions and avoids vendor lock-in: a situation where the provider raises prices, restricts functionality, or ceases operations in a needed jurisdiction.

The irony of Palantir's position is that the company itself is a publicly traded corporation focused on contract growth. Its arguments for sovereignty are equally commercial; they're simply monetized differently: not through tokens, but through licenses, implementation, and support.

What this means

Palantir's manifesto marks a growing divide in the AI industry: cloud model versus sovereign. As AI becomes embedded in the critical infrastructure of states and corporations, this dispute will only intensify. Palantir is betting that large institutional clients will ultimately choose control over technology rather than the convenience of cloud access.