The loudest AI risk debate is about a hypothetical future: a superintelligence that outsmarts humanity. It’s a worthwhile argument. It also conveniently skips the harm that needs no speculation at all, because it has already happened.

People have been wrongfully arrested when facial recognition matched the wrong face. Automated systems have denied benefits, loans, and jobs in ways that turned out to be biased — and opaque enough that the victims couldn’t even see why. Fraud using cloned voices and faces is now routine.

None of this required a sci-fi breakthrough. It required ordinary software, deployed at scale, trusted more than it deserved, with no one accountable when it failed. That’s the danger that’s measurable today — and the one this blog tracks, case by case, source by source.