Something very important is happening this week in AI news, and we need to pay attention to it.
On Tuesday February 24th, the founder and CEO of Anthropic was summoned to the Pentagon to meet with Pete Hegseth, the US Secretary of Defense, for a "tense meeting."
Anthropic, for context, is under contract with the Pentagon for use of its AI assistant Claude in US military operations. (It was recently revealed that Claude was used in the kidnapping of Venezuelan president Nicolás Maduro.)
Anthropic was given an ultimatum: allow unrestricted military use of Claude—including surveillance and autonomous weapons—or risk being blacklisted as a “supply chain risk.” Anthropic says they refused. Now we wait and see what the Pentagon responds.
But it's important to take a step back and look at what this crisis reveals about AI today.
The AI debate is often framed as being about “regulation”—as if the right rules will make it “safe.” But this framing obscures what we, and many others believe, is the ultimate function of AI—that it's a tool built for control. To control media, to control culture, to control the conversation.
AI is infrastructure for power.
The companies racing to build it now are under immense pressure to win—to be the one AI company to rule them all. And that means agreeing to the terms of the highest bidder.
We know that Google, OpenAI, Musk's xAI, et al are all more than happy to do that. (They've all donated lavishly to Trump as it is.) And as long as one of these companies is willing to take the low-road—to forgo ethics, to break laws—an AI-safe world is impossible.
In other words, whether or not Anthropic complies doesn't really make a difference, because another AI company—with the same or equal capabilities—will. It doesn't matter who wins the race because, in the end, it's a rigged game.
Which is why when it comes to creating an AI-safe world, the solution isn't to choose the "good, ethical" AI company over the “bad” ones.
The only truly AI-safe world is a world with AI-free spaces and institutions.
