Palantir's Karp bashes OpenAI, Anthropic token model: 'Something has gone completely wrong'

Palantir CEO Alex Karp delivered a blistering critique of frontier AI laboratories on Wednesday, accusing OpenAI and Anthropic of running an "effing insane" business model that leaves enterprises paying escalating token costs for limited value while risking their proprietary data and intellectual property, per CNBC, which carried the remarks under the banner "Something has gone completely wrong."

Mr Karp said the top AI labs were misleading corporate partners — "overselling" the risks of AI while at the same time offering their most powerful models to companies and governments worldwide, per CNBC's account of the interview. ZeroHedge also carried the remarks.

The tenor of the exchange was itself part of the story: when CNBC host Becky Quick observed, "You sound pretty angry," Mr Karp did not demur, launching into a reply invoking the voice of America's builders, per the account. Neither OpenAI nor Anthropic's response is carried in the material supplied.

Background

Palantir, co-founded by Mr Karp in 2003 and known for data-analytics work with defence, intelligence and government customers, has positioned its Artificial Intelligence Platform as the layer that makes large language models useful and safe inside real enterprises — a pitch that sets it up as both customer and rival of the frontier labs whose models it orchestrates. Mr Karp is among the most quotable and combative chief executives in technology, and broadsides delivered on financial television are a familiar part of his repertoire.

The economic argument beneath the rhetoric is a live one across the industry. Frontier labs sell intelligence by the token — metered units of model input and output — and enterprise customers have increasingly questioned whether metered costs that scale with usage deliver commensurate business value, and whether feeding proprietary data into external models exposes their intellectual property. The labs, for their part, have built rapidly growing enterprise businesses on exactly that model, making the dispute a contest over where the profits of enterprise AI will settle.

What comes next

Neither OpenAI nor Anthropic had responded in the material reviewed, and whether they engage publicly is the immediate thing to watch. The more consequential measure will be commercial rather than rhetorical: enterprise renewal rates, token pricing and the competitive positioning of platform vendors like Palantir against the labs' own enterprise offerings in the quarters ahead.