Palantir CEO Alex Karp criticizes AI model “theft” of enterprise value, argues frontier labs benefit more than customers

Palantir CEO Alex Karp used a recent CNBC appearance to sharpen a long-running critique of how frontier AI model providers interact with enterprises, arguing that customers deploying leading systems from firms such as OpenAI and Anthropic often feel they are giving up too much of their own business value.

According to coverage of the interview, Karp characterized many enterprises as “livid” about what he described as AI models that effectively “steal” the value generated by a company’s data, workflows, and business context. The thrust of his argument was that while model vendors capture outsized returns from the ecosystem, enterprises bear the costs of integrating AI into operations and organizing their proprietary information so the models can perform.

Karp’s remarks frame Palantir’s business pitch as a counterweight to that dynamic. Palantir sells software that is designed to bring together data and make it operational inside organizations, rather than treating AI simply as a black-box service to be prompted. The underlying message, as reflected in the interview coverage, is that customers should be able to preserve the benefits they create from their own business intelligence, not see that advantage leak away when they adopt widely available foundation models.

The criticism also lands in a broader debate about how AI value is distributed across the stack. Model providers sit at the center of the modern AI economy, but enterprises provide the domain knowledge, operational constraints, and proprietary datasets that make AI useful beyond generic demonstrations. When those enterprise inputs are not reflected in pricing, licensing, or deployment arrangements, executives sometimes argue that the result is an asymmetry in bargaining power.

Karp’s choice of language, including the claim that models can take away “business value,” is notable because it goes beyond general concerns about competition and into the mechanics of value creation. In the CNBC-linked coverage, he suggested that the people doing the work of translating business processes into AI-ready systems do not always receive a commensurate share of the upside.

Palantir, which trades on the Nasdaq under the ticker PLTR, is already closely associated with enterprise customers in government and commercial settings. Its software is commonly described in the market as an integration and deployment platform for data-driven decision making, and the company positions itself as enabling organizations to operationalize analytics and AI in ways that align with their internal governance and operational needs. Karp’s latest comments are consistent with that positioning, emphasizing control, operationalization, and enterprise ownership of outcomes.

What the interview coverage does not provide, at least in the information available here, are specific contractual details about which arrangements Karp believes are most problematic, whether he is referring to particular pricing terms, or whether he expects model providers to change licensing or disclosure practices. The remarks are also not accompanied in the reported material by quantified estimates of the value transfer he is describing.

For investors and industry watchers, the immediate takeaway is that Palantir’s CEO is pressing a narrative that could resonate during enterprise AI procurement cycles. The next thing to watch is whether Palantir elaborates on this theme in subsequent commentary, including whether it will tie the critique to measurable outcomes such as adoption rates, customer retention, or how Palantir’s platform is used to reduce perceived “value leakage” when enterprises deploy external AI models.