Palantir CEO Alex Karp Calls Out OpenAI and Anthropic as Overcharging, Data-Stealing, and a National Security Risk

What Karp Actually Said

Alex Karp did not wander into polite criticism. On CNBC's Squawk Box on July 1, the Palantir CEO called the business model of companies like OpenAI and Anthropic "effing insane" and said enterprise leaders across the country are privately "livid."

His core accusation, as reported by Forbes and Business Insider: frontier AI labs double-dip. They charge companies escalating fees for token-based access, then use those same companies' data to improve their own models. Karp called it a "wealth tax." Businesses pay for a service while simultaneously surrendering the competitive edge that makes them worth anything.

"The basic view among enterprises in this country is 'I'm going to chillax and waste my time with tokens, I'm going to get no value, and they're going to get my IP,'" Karp told Squawk Box anchor Becky Quick, according to Business Insider.

When Quick observed, "You sound pretty angry," Karp replied: "This is the voice of American business that is being channeled through me."

The Token Model Under Fire

The technical complaint is specific. According to Tom's Hardware's coverage, when Karp talks about AI companies stealing customers' "weights and alpha," he means the structural relationships between a company's data and the business processes built on top of it. This is the operational blueprint of a business.

Karp posed a challenge that no frontier lab has answered publicly: if AI tools generate the enormous value vendors claim, why don't they charge a percentage of that value instead of a flat token fee? Palantir CTO Shyam Sankar made the same point more bluntly earlier, saying "more tokens means more slop" — a direct shot at what the industry has been calling "tokenmaxxing."

Palantir's own business model is structured specifically to counter this. Tom's Hardware notes that many of Palantir's products run on-premises or in sovereign cloud environments, carry DOD-required CMMC Level 2 certifications, and the company says it does not retrain models using customer data. Karp is not a neutral observer here. He's pitching an alternative.

National Security Is the Sharper Edge

The data-fee argument is a business dispute. The national security argument is harder to dismiss.

Karp told CNBC: "Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane." He appeared to be referencing, at least in part, Anthropic's refusal to remove safeguards that would allow its models to be used for mass domestic surveillance or fully autonomous weapons. That position led the Pentagon to designate Anthropic a "supply chain risk" in March, according to Forbes.

Days before that designation, the Pentagon reached a separate deal with OpenAI that drew criticism from AI policy and legal experts, Forbes reported. President Trump signed an executive order in June requiring federal oversight of new AI models before public release. OpenAI has said broader access to its newest models will come after a "limited preview" period.

Karp appeared on Squawk Box specifically to discuss Palantir's new partnership with Nvidia, called the Sovereign AI OS Architecture, designed to help the U.S. government use advanced AI in secure environments.

The Strongest Counterargument

OpenAI and Anthropic have not been given a fair pass here, but their position deserves clear statement before judgment.

Both companies argue that their model terms of service prohibit using customer data to train models without consent. Anthropic specifically has positioned its restrictions on military use as a safety feature, not a commercial shakedown. The company has said it won't let its models be used for weapons that remove meaningful human control. That's a principled position, whatever one thinks of it. The claim that frontier labs are actively and covertly harvesting enterprise data has not been established by any regulator, court, or named investigator. No charges have been filed. Karp's accusation is pointed, but it is an allegation, not a proven fact.

Microsoft CEO Satya Nadella raised a parallel concern in a more measured way earlier this month, according to Business Insider. He worried that entire industries could have "their knowledge commoditized right out from underneath them." That's a legitimate systemic worry about AI economics, not necessarily evidence of intentional data theft.

What Moves the Market

Palantir shares rose approximately 9% on July 1, the day of the interview, according to both Forbes and Tom's Hardware. Shares of other AI companies saw a dip, Tom's Hardware noted, though specific figures were not reported.

Karp's net worth sat at $12.3 billion as of July 1 per Forbes estimates. He co-founded Palantir with Peter Thiel ($27.4 billion) and Stephen Cohen ($4.6 billion). The company went public via direct listing on the NYSE in 2020.

The Open Question

Palantir posted a nine-point manifesto on "AI sovereignty" around the same time as the interview, per Business Insider. Point one: "Data retention is your treasure. Transfer it at your own peril." It's a marketing document as much as a philosophical one.

The unresolved question sitting underneath all of this is whether OpenAI and Anthropic are actually retaining and using enterprise customer data in ways their contracts prohibit, or whether Karp is making a competitive argument dressed up as a public interest warning. That distinction matters enormously. As of July 3, 2026, no regulator has publicly investigated the specific data-harvesting practices Karp described.