Palantir's Karp Escalates Fight With OpenAI and Anthropic
TL;DR
- Palantir CEO Alex Karp told CNBC frontier AI labs are 'trying to drug addict us to a future they believe they control.'
- Karp rejected the China-distillation framing, saying frontier labs themselves 'distilled all the value of IP, everywhere.'
- In late July, Palantir joined Nvidia, Microsoft and others on an open letter defending open-weight AI models as essential for national security.
Alex Karp used Palantir's earnings-day interview cycle on CNBC to swing back at OpenAI and Anthropic, and this time he did not bother with hedges. According to CNBC's reporting, the Palantir CEO said frontier AI labs are 'trying to drug addict us to a future they believe they control,' a line that lands harder than his July complaint that 'something has gone completely wrong' with the token-based pricing model. Second time in about a month, same venue, same targets. The pattern matters more than any single quote.
The interesting move underneath the rhetoric is how Karp reframes the distillation debate. Washington has spent the last year framing Chinese labs distilling U.S. frontier models as the threat. Karp inverts it, telling CNBC that Chinese models can't be blamed when the frontier labs themselves 'distilled all the value of IP, everywhere.' Translated into enterprise buyer language, he is telling CIOs that sending proprietary data through OpenAI or Anthropic endpoints is itself the leak, and that keeping the weights inside the firewall is the fix.
That framing is not free-floating. In late July, Palantir signed an open letter alongside Nvidia, Microsoft and other companies defending open-weight AI models as essential for national security. Open weights are downloadable and modifiable, and CNBC notes they can perform similar tasks at a fraction of the price. Palantir has a clear commercial reason to want that door held open: its AIP and Foundry platforms are natural wrappers around whichever open-weight model a regulated buyer picks, and a Sovereign AI reference stack with Nvidia's Nemotron models is already on the table.
The honest caveat is that a CEO with an Nvidia co-marketing motion and an interest in commoditising his suppliers is not a neutral witness. What the reporting does not give you is the concrete customer evidence: no named enterprise migration off OpenAI or Anthropic, no numbers on the cost delta inside Palantir's own accounts, and no clarification of what 'distilled all the value of IP, everywhere' is actually alleging in legal terms. Take the specifics as reported, not settled.
The forward-looking piece is who gets to use this. If Karp's line lands with defense, financial and critical-infrastructure buyers already nervous about IP exposure, it becomes a procurement talking point that pushes frontier labs toward lower pricing, tighter data-handling guarantees, or their own open-weight releases. That is the shift worth watching over the next few earnings cycles, not the insult itself.
Originally reported by cnbc.com
Read the original article →Original headline: Palantir's Karp Renews Attack on OpenAI and Anthropic: Frontier Labs Are 'Trying to Drug Addict Us'