Comment by gleipnircode

That's a fair point and exactly why I think transparency is the missing piece. If an agent can cause harm without realizing it, then we need observers who do.

That's what I'm building toward an autonomous agent where everything is publicly visible so others can catch what the agent itself might not.