Palantir (PLTR) CEO Alex Karp has dropped a blunt theory about the AI industry’s safety push.
The outspoken AI executive said that some companies might need government ownership to escape the financial consequences of their technology.
In his CNBC interview, Karp framed nationalization as the destination he believes those businesses are moving toward, with liability protection the prize.
His startling comments come as the industry’s conversation shifts towards safeguards and whether development should slow down. Moreover, that also raises the question of who shoulders the risks created along the way?
Karp zeroes in on customers’ intellectual property. He said businesses working with Palantir believe that their ideas and data are being fed into AI models, putting their competitive advantages at risk.
That allegation also serves Palantir’s commercial pitch around helping customers retain control of their information, something Karp has doubled down on in recent months.
During the company’s Q2 2026 earnings call, Karp cautioned businesses relying too much on external AI providers, noting that it could ultimately help them build “a competitive business that doesn’t require your business or your people.”
Nevertheless, AI developers want businesses to trust their systems, while Karp is questioning whose interests the proposed safety measures would ultimately protect.
Karp says AI’s safety push could become a liability shield
Karp makes the case that AI companies could seek nationalization to contain liabilities stemming from their technology and their use of customers’ intellectual property.
More AI:
- Nvidia just made a move Wall Street wasn’t ready for
- Microsoft just took sides in AI policy fight
- OpenAI just disclosed something genuinely alarming
“It has to be nationalized,” he told CNBC, describing what he believes their business model requires.
His reasoning starts with corporate customers. Karp said Palantir’s clients are “livid” about what they believe is “the theft of their alpha”, which is the proprietary knowledge that gives their businesses a competitive edge.
He alleged that discounted AI tokens help providers acquire that knowledge: “They’re discounted so that they can get access to your IP so that their models get better.”
Put simply, customers could help improve systems that eventually benefit their rivals, undermining the advantage they were trying to build.
Karp then connected those concerns to lawsuits, saying, “Every single one of my clients is going to sue” if the businesses are not nationalized.
The clear takeaway is that stronger safety rules could increase providers’ accountability. Liability protection could reduce their exposure. Karp’s claim is that the industry wants the latter wrapped inside the former.
AI leaders back safeguards, but differ on slowing development
The support for safety measures isn’t exactly a uniform commitment to stop building AI.
- Anthropic: Dario Amodei advocates slower capability advances, embedded independent evaluators, and international coordination. His proposal explicitly allows continued training.
- OpenAI: The Verge reports that, Sam Altman supports pacing development and independent evaluators, while distinguishing slower progress from stopping altogether.
- Google DeepMind: That same Verge piece said that Demis Hassabis endorsed Amodei’s direction and pointed toward common AI standards.
- Microsoft: Satya Nadella offered qualified support, advocating a broader pool of evaluators.
- Amazon: Reuters reports that the company called for rigorous testing and safeguards without endorsing an industry-wide slowdown, illustrating the distinction between stronger oversight and slower development.
Karp and Nadella want businesses to own their AI knowledge
It’s important to note that Karp’s sovereignty pitch has been building for months.
In July, he was already criticizing token economics and arguing that customers should control their data. September’s AIPCon made that pitch more concrete, with expanded ties to Nvidia and locally hosted AI offerings designed around proprietary information.
Microsoft CEO Satya Nadella identified a similar tension in his July 12 essay: “You essentially pay for intelligence twice, once with money.” The second payment, he explained, is the business knowledge needed to make AI useful.
His proposed answer included ownership of feedback and institutional memory, private learning environments, and the ability to change models without losing accumulated expertise.
Put simply, whoever controls those workflows and learning processes can become harder to replace than the underlying model supplier.
For Karp, that supports Palantir’s role connecting AI to business operations. For Nadella, it supports enterprise infrastructure that preserves customer control.
Both argue that durable AI value belongs closer to the customer’s business. Their companies stand to benefit by supplying the systems that keep it there.
Palantir’s valuation leaves little room for disappointment
Karp’s sovereignty might strengthen Palantir’s sales pitch, but its valuation already demands exceptional execution.
Seeking Alpha data puts Palantir at 108.46 times forward non-GAAP earnings, versus a sector median of 22.41. Moreover, it’s changing hands at a forward sales multiple of 51.17, compared with 3.35 for the sector, which means investors are paying heavily for anticipated growth.
The historical comparison is revealing. Palantir’s forward earnings multiple sits about 14% below its five-year average, yet its forward sales multiple remains roughly 43% above its historical average.
That combination suggests investors expect stronger profitability to help justify the revenue premium. A lower earnings multiple alone does not establish that shares are cheap.
The operating opportunity is clear, but shareholders need evidence in contract expansion, customer retention, and cash generation per share. Rival platforms offering similar protections could constrain pricing.
Even strong growth may disappoint investors if valuation compresses. That said, Palantir must convert its sovereignty pitch into results that exceed expectations.
