When Your AI Vendor Picks a Side, So Do You
Three AI companies faced the same question. Three different answers. What's yours?
Roughly 10 days ago, Palantir posted a thread on X. One thousand words. Twenty-two points. Thirty-two million views in forty-eight hours.
If you haven’t read it yet, you’re probably in a minority. And if you did read it, or even just the headlines about it, you likely had a reaction. Strong enough to share, or strong enough to close the tab quickly.
That reaction is exactly where I want to start.
Because the Palantir manifesto isn’t really about Palantir. It’s a mirror. It forces a question that most leaders in tech have been quietly avoiding for years: what do the companies building your AI infrastructure actually believe and what will they do when those beliefs get tested?
The manifesto made it impossible to look away. But the real story started couple of months earlier.
What Actually Happened in February
Let me give you the compressed version, because it’s important.
The Pentagon wanted Anthropic, the company that makes Claude, to grant the US military unrestricted access to its AI models for any lawful purpose. Anthropic said no. Specifically, the company drew two hard lines: no use for fully autonomous weapons, and no use for mass domestic surveillance. The talks collapsed.
On February 27, President Trump ordered all federal agencies to immediately cease using Anthropic’s technology. The Department of Defense designated Anthropic a “supply chain risk to national security”, a label historically reserved for foreign adversaries. Anthropic became the first American company to receive it.
The same day, OpenAI signed a deal with the Pentagon.
Then, on April 19, Palantir posted the manifesto. Its stock barely moved, down 0.34%. The market read it not as a liability, but as a conviction. Which tells you something.
Three companies. Three very different answers to the same question: at what price do you hand over your technology to be used in war?
Three Answers, Three Identities
I want to be careful here, because it’s easy to turn this into a morality play with heroes and villains. That would make for a satisfying read and a lousy analysis.
Palantir didn’t need to negotiate with the Pentagon. Its position was already baked into its identity and the manifesto simply said it out loud. “The question is not whether AI weapons will be built,” it states, “it is who will build them and for what purpose. Our adversaries will not pause to indulge in theatrical debates.” Palantir has a $10 billion contract with the US Army. The manifesto isn’t a pivot. It’s a confession of what was already true. The reason it shocked so many people is that most tech companies hold the same positions but have the good sense, or the PR instinct, not to publish them.
Anthropic said no and paid for it. Blacklisted, sued, faced with the threat of the Defense Production Act being invoked against it. The company held its red lines on autonomous weapons and surveillance, even as its own safety policies were under pressure from competitive dynamics. It’s worth noting the bitter irony: while the Pentagon was publicly designating Anthropic a national security threat, the NSA was quietly using Anthropic’s newest, most capable model, Mythos, for cybersecurity work. The US government was suing its vendor and using its product simultaneously.
OpenAI signed the deal. The same day Anthropic was frozen out. Its CEO later admitted the initial terms had been “opportunistic and sloppy,” and the company published an updated version of the contract terms, claiming its safeguards were stronger and more enforceable than what Anthropic had originally negotiated. Maybe they are. But the timing has its own logic. OpenAI had spent the previous year hiring over a dozen former defense officials and lobbyists. It was ready for the moment Anthropic wasn’t.
None of these companies are purely good or purely cynical. All three are navigating the same impossible tension: you’re building something powerful enough to be strategically important, and powerful enough to be dangerous. The question is who gets to decide where the line is.
Why This Is Your Problem Too
Think about your organization for a moment. You’re probably not building weapons. You’re not working with governments. You’re using AI for documentation, code review, customer support, internal workflows, the standard enterprise stuff.
And yet the infrastructure you depend on is built by companies whose values, politics, and financial pressures are not yours to control. Those values don’t stay in the background. They surface, in contract negotiations with governments, in decisions about what the technology will and won’t do, in public manifestos that get 32 million views overnight.
Eliot Higgins, founder of the investigative platform Bellingcat, put it precisely when the manifesto dropped: “These 22 points aren’t philosophy floating in space. They’re the public ideology of a company whose revenue depends on the politics it’s advocating.”
He was talking about Palantir. But the observation applies to all three.
Every AI company has an ideology. Palantir’s is explicit, admirably or disturbingly so, depending on your perspective. Anthropic’s is embedded in its corporate structure: it’s a public benefit company controlled by a trust whose mandate is safe AI development, not profit maximization. OpenAI’s is increasingly shaped by its race toward IPO and a $30 billion revenue run rate.
When you choose a vendor, you’re not just choosing a model. You’re choosing whose version of “responsible” you’re betting on.
The Question Leaders Aren’t Asking
I’ve sat in enough strategy meetings to know that most organizations choose their tools (AI or anything else) based on three criteria: performance, price, and integration effort. Occasionally, data privacy regulations make it into the conversation.
What almost never comes up: what does this company believe, and what will it do when belief gets tested?
February 2026 was a test. All three companies got tested, all three gave different answers, and the consequences were real for them, and for every organization that depended on their infrastructure.
I’m not suggesting you need to audit the geopolitics of every software vendor you use. That’s not practical and, frankly, it’s not your job.
But I do think there’s a more honest version of the question worth asking, especially for leaders in European organizations increasingly dependent on American AI infrastructure: do you know what your vendor believes, and are you comfortable with it?
That’s the question February answered for three companies. It’s worth asking before someone answers it for you.
What happened these past months opened a space that didn’t exist before. Most organizations would prefer not to take a position and that’s understandable. But the space is open now. The next time this kind of pressure lands on a company whose tools you depend on, “I hadn’t thought about it” won’t hold up as an answer anymore.
What’s your take? Does this kind of vendor values question belong on your leadership agenda, or is it too abstract to be useful? I’d genuinely like to know how you’re thinking about it.
Reply here or find me on LinkedIn. The conversation is always more interesting than the article.