Picture your gym trainer copying your workout routine, packaging it up, and selling it to the guy on the next treadmill — while still charging you full membership. That's basically what Alex Karp just accused the entire AI industry of doing, live, on CNBC, with the kind of volume usually reserved for family arguments at Thanksgiving. The Palantir CEO didn't do a polite fireside chat. He did a 20-minute, red-faced monologue that had the anchors visibly bracing themselves. And buried in the chaos was a genuinely sharp point about who's actually winning the AI gold rush (spoiler: probably not you).

Here's what actually happened, once you strip away the theatrics. Karp went on CNBC to talk about Palantir's expanded Nvidia partnership and instead used most of the airtime to torch the business model of OpenAI, Anthropic, and the rest of the frontier AI pack. His core claim: enterprises are paying escalating fees for AI "tokens," getting underwhelming returns, and — worse — handing over their proprietary data and competitive edge to the very companies they're paying. He called it a "wealth tax." He said the whole setup was "effing insane." He also floated, somewhat dramatically, that he might "get kicked out of the room" for saying it.

The part worth actually paying attention to: Karp claims the CEOs he talks to privately are furious, even if they'd never say so on the record. Companies are quietly paying premium prices for AI tools while suspecting (or knowing) their prompts, workflows, and business strategy are training the next version of someone else's model. That's the landlord-with-a-spare-key problem — you're renting the space, but someone else still has access to everything inside it, and they're using what they see to furnish the unit next door for your competitor.

Karp's pitch, of course, isn't purely altruistic (he is, after all, selling an alternative — Palantir's "Ontology" layer, which promises to let businesses use AI without exposing sensitive data). But the underlying frustration he's tapping into is real, and it's spreading. Several major companies have already started shifting workloads toward lower-cost, open-weight models specifically to dodge the token-cost-and-data-exposure combo Karp was ranting about.

If you're running a business — SME, startup, department, doesn't matter — the AI spending line on your P&L deserves a harder look than it's probably getting. It's easy to sign up for a shiny new AI tool because everyone else has one (peer pressure doesn't stop at age 12, apparently). But Karp's underlying question is the one you should be asking your vendors directly: where does our data go, who trains on it, and what happens to our competitive advantage once it's in someone else's system? These aren't paranoid questions anymore; they're basic procurement hygiene.

This also matters because the fight isn't staying in boardrooms. It's spilling into national security territory too — the Pentagon has already flagged one major AI lab as a "supply-chain risk," and the politics around AI vendor trust are shifting fast. When the loudest voices in the room start questioning whether frontier labs can be trusted with sensitive data, "just trust us" stops being a viable sales pitch, for governments and for you.

The practical takeaway: don't cancel your AI subscriptions in a panic. Do start asking sharper questions about data ownership, model training rights, and what you're actually getting for the money — before the next monthly invoice lands and you're still not sure who benefited more from the relationship.

Karp may have delivered his take like a man auditioning for a courtroom drama, but the question underneath the shouting is one every business owner should be losing a little sleep over: are you the customer here, or the product? Somewhere between "effing insane" and "livid," there's a real conversation happening about who owns the value AI creates — and right now, it's not entirely clear the answer is you.

— The Business Index Team