'Effing insane': Palantir CEO Alex Karp attacks OpenAI, Anthropic over token-based AI model

Palantir CEO Alex Karp has taken direct aim at some of the biggest names in artificial intelligence, criticising the token-based business models used by OpenAI and Anthropic while arguing that governments and enterprises should have greater authority over the AI systems they deploy.

Speaking in an interview with CNBC, Karp described aspects of the AI industry as "effing insane", accusing leading model developers of creating an economic structure that, in his view, delivers limited value to customers while encouraging dependence on AI providers.

His remarks underscore a growing divide within the AI sector over pricing, ownership of AI models and the role of private companies in supplying technology for national security and critical infrastructure.

Karp questions token pricing and enterprise value

Most frontier AI companies, including OpenAI and Anthropic, charge customers based on token usage. Tokens are the units of text that AI models process when analysing prompts or generating responses, making them the foundation of how many commercial AI services are priced.

More from Tech

Karp suggested that this approach has failed to produce meaningful returns for businesses despite increasing spending on AI.

"So the general way these things were sold... Sam (Altman, OpenAI cofounder) and Dario (Amodei, Anthropic co-founder), there's nothing more fun than debating Dario in private. So I'm not throwing shade at them, but something has gone completely wrong," Karp said.

He argued that many corporate customers believe they are paying for AI usage without receiving proportionate business benefits, while also risking exposure of valuable intellectual property.

"And 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 (Intellectual Property),'" he said.

Karp went further, claiming that many executives have become increasingly frustrated with AI vendors.

"Every single enterprise I deal with... these people are livid. They're like, 'I am paying for tokens that create no value. These people are stealing the weights and alpha of my business,'" he said.

Palantir, which develops AI-powered software for governments, defence agencies and large businesses, has increasingly promoted its own approach as one that allows organisations to retain greater control over their data and AI deployments rather than relying entirely on model developers.

AI control becomes another battleground

Beyond pricing, Karp also challenged the idea that AI companies should retain primary control over the underlying models used in sensitive government and military applications.

During the interview, he questioned whether defence agencies should depend on commercial AI companies to determine how models are deployed in national security operations.

"In the classified context, when the Department of War goes to you and says, 'I need this application,' do they get to control the weights to do it? Or do you get to control the weights? Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane," Karp said.

In AI systems, model "weights" are the internal parameters that determine how a model behaves and can be adjusted or fine-tuned for specialised tasks. Karp argued that organisations using AI, particularly governments, should retain authority over those systems rather than leaving such control solely with AI developers.

His comments also come amid growing tensions between Palantir and Anthropic over the use of AI in defence. Anthropic has positioned itself as a company that prioritises safeguards around advanced AI deployment and has resisted allowing its technology to be used for applications such as mass surveillance and autonomous weapons. That stance has reportedly contributed to friction with the Trump administration, with the Pentagon reducing contracts involving the company.

Palantir, meanwhile, has expanded support for alternative AI providers, including Elon Musk's xAI, to ensure continued access to advanced models for defence customers.

Karp framed his criticism as reflecting concerns shared by corporate America rather than simply representing Palantir's commercial interests.

"This is the voice of American business that is being channelled through me," he said, adding that enterprises were "tired" of what he characterised as AI technologies being "oversold".

His remarks highlight a broader debate that is likely to intensify as AI adoption accelerates: whether the future of artificial intelligence should be shaped primarily by model developers, or by the governments and enterprises that increasingly depend on the technology.