Palantir's Alex Karp Warns AI Could Predict Your Thoughts. Color Me Skeptical.
Palantir CEO Alex Karp says AI is about to jump from guessing what you'll buy to guessing what you'll think. The warning lands while his own stock rides a historic run. Here's why the risk is real, the framing is marketing, and the difference matters.
Alex Karp says AI is about to make the leap from predicting what you'll buy to predicting what you'll think. Color me skeptical, but not for the reason you'd guess.
The Palantir CEO has been making the rounds with a warning that sounds like science fiction with a deadline attached. Today's models guess your next purchase. Tomorrow's, he suggests, might guess your next opinion. Palantir shares climbed roughly 340% in 2024, and the company's market cap has pushed north of $250 billion at points in early 2025. People are listening. They should, even if they should also check who's talking.
So what's actually behind the claim?
The Run That Made Him Loud
Start with the receipts, because they explain the volume. Palantir joined the S&P 500 in September 2024. Its fourth quarter 2024 revenue hit $828 million, up 36% year over year, and management guided 2025 revenue to roughly $3.75 billion. US commercial revenue climbed more than 60%. That's a real business, not a story stock, and Karp knows a megaphone when he's holding one.
His thesis is straightforward enough. Recommendation engines already nail your shopping habits with unsettling accuracy. Proponents of the next wave argue the same math works on intent, on belief, on the stuff you'd rather keep in your own head. If a model can infer your politics from your search history, what stops it from inferring your doubts?
The question worth asking: is that a prediction, or a very good guess dressed up as one?
Where The Argument Gets Thin
Here's the counterpoint, and it's a serious one. AI doesn't read minds. It finds correlations in data you or somebody else already handed over. There's no pipeline from your skull to a server rack. Admittedly, that distinction matters less to you if the inference lands often enough to shape what you see, read, and buy.
And be fair to the skeptics on the other side of this. Karp runs a company whose core product fuses data for governments and militaries. His track record on privacy warnings is mixed, to put it gently. When the guy selling the monitoring tools warns you about monitoring, a little suspicion isn't paranoia.
Remember Cambridge Analytica? In 2018 that firm claimed it could model your personality from Facebook likes and swing elections with it. Later reporting found the effect was far smaller than the pitch. The narrative outran the evidence by a mile.
Regulators aren't sitting still either. The EU AI Act's first obligations took effect in 2024, and several US states passed their own rules around the same window. History suggests the loudest alarms often sound right before the compliance contracts get signed.
My Verdict
Karp is describing a real risk and monetizing the fear of it at the same time. Both things can be true. I'm not entirely convinced the thought-reading scenario arrives on schedule, but I'm convinced the inference layer gets sharper every quarter, whether anyone legislates it or not. That's the part worth worrying about.
Watch three things from here. Palantir's government contract wins, because they show where the capability lands first. State and federal privacy bills in 2025, because they'll decide whether you ever get to audit the guess. And Palantir's own commercial AI products, because if thought prediction is coming, its biggest vendor will show up early and charge for it.
Time will tell, though. It usually does, just later than the headlines promise.
