Palantir's CEO Slams OpenAI and Anthropic: Who Actually Controls the AI?

Palantir's CEO (as characterized in this commentary) argues the most glaring gap in today's AI industry is that no one is clear on who controls the models, who holds the weights, and who captures the profit. Paying token fees indefinitely leaves customers with no lasting asset -- and worse, their trade secrets and know-how risk being absorbed by the AI provider itself.

For government agencies or mission-critical corporate operations, he argues, using AI without absolute clarity on where the data ends up is simply too dangerous. Organizations should be using AI under their own control, not someone else's.

Deploying AI in environments with strict operational rules requires something above the model itself -- an 'application layer' that connects AI to the ground truth: pulling in the right data and deciding how to judge it. The NVIDIA partnership is framed as a move to change exactly that dynamic.

That partnership, he says, hands computing control back to customers -- keeping the model, the data, and above all the competitive know-how that matters most out of anyone else's hands. He also emphasizes a 'model-agnostic' philosophy: customers should always be free to switch to whichever AI model they prefer, without being locked into any single vendor.

The discussion turns to what he sees as a glaring contradiction: frontier AI labs market their models as available to virtually anyone, yet hesitate to make them available to their own country's defense department. Being able to serve adversarial actors while withholding from the military protecting one's own nation raises the question of whether that model can genuinely be called 'safe.'

What's actually needed, he argues, is an absolute guarantee that a company's confidential information and know-how won't leak externally -- and no organization can afford to hand over the core of its business to a provider that can't offer that guarantee. This isn't framed as one company's opinion but as a matter touching the interests of an entire industry, even national assets.