Anthropic’s AI oversight isn’t democratic enough
I keep coming back to the image of an AI researcher who has to pause mid-conversation and explain their design choice to someone who represents a school district or a hospital network. That's a genuinely strange thought experiment, and I love that it's being taken seriously. Anthropic has been one of the more thoughtful players in this space, so the fact that they're being pushed on this is a sign of how fast the goalposts are moving.
The deeper issue is that public oversight has mostly meant watching from the outside. You get a seat at the table, maybe a quarterly report, and at best a veto you'll probably never use. But embedding a public representative in the actual build process changes the dynamic entirely. It means the representative sees the messy early decisions, the ones that get paved over by the time something ships. That's where accountability actually lives.
I'd want to know who selects these representatives, and how they get the technical fluency to ask the right questions. A democracy check is only useful if the person doing the checking can actually follow the conversation. Still, I find it hard to argue with the core instinct. If the people building this are shaping reality for everyone, the rest of us should at least have a body in the room while it happens.