Original Title: White House officials reportedly frustrated by Anthropic’s law enforcement AI limits

But that doesn't fit here so I shortened it the best I could.

I'm frankly surprised they didn't build in some ability to flexibly apply guardrails. You'd have to think a product manager brought up the desirability to have variably locked down systems for different end users, this being an obvious application.

Did people decide that they'd design in lines that can't be crossed and everyone just went with it? Must have made for some interesting conversations.

Extraordinary that Anthropic has shown some detectable backbone

I know very little, I had assumed they had the same level of subsequent obedience of the other "tech bros"

Wrong, it seems