The AI That Kills and Nobody Knows How: Anthropic's Black Hole

Here's a problem everyone's pretending not to see: the people who create insanely powerful AI have no idea how the military actually uses it when they sell it to them. Dario Amodei, Anthropic's CEO, had to publicly admit he doesn't know if Claude — his AI model — was involved in coordinating or planning an aerial strike that hit a school in Iran.

This isn't a small thing. We're talking about 120 dead children. And the answer from whoever created the tool possibly used in the attack is: "Beats me, I don't know." It's like a car manufacturer selling vehicles to criminals and then saying "we have no idea if our cars were used in robberies."

The real scandal is that Anthropic, like most AI giants, sells its most advanced tools to the US Department of Defense without having any actual control over how they end up inside military systems. No transparency, no traceability, nothing. Just money coming in and AI going out to do who-knows-what.

Amodei admitted there's a "broader knowledge gap" between what CEOs know and what the military does with their products. Translation: "We have no clue what's happening down there, and honestly we didn't try very hard to find out." It's almost admirably honest, if we weren't talking about death and technology.

The whole thing highlights an absurd paradox of our moment: the companies building the smartest machines on the planet remain completely in the dark about how they're used to decide who dies. It's like creating the most precise rifle ever made, selling it to the military, and then saying "we don't know where it shoots, fingers crossed."

This isn't just an Anthropic problem. It's systemic. Every big tech company selling AI to the military is in the same boat: aware their software might be used in lethal operations, but conveniently clueless about the details. It's almost perfect, if you want to make money without feeling responsible.