This 'College Protester' Isn't Real. It's an AI-Powered Undercover Bot for Cops

I wonder how much of this will just encourage protests and radicalization. If your agent is trained to match a profile of a radical, then it necessarily is spreading and encouraging that radical messaging in order to fit in and gain trust. At least with real agents there is a plausible mechanism for their judgement to filter out who is targeted and they can't infinitely propagate like the AI could.

It's already somewhat normal for cops to try and radicalize people to create evidence for arrests so it's only a question of scaling up, right?

That's what I worded it the way I did. At least theoretically the humans have some judgement that could limit who they go after or how far they push and can't propagate infinitely. The issue with AI is the even greater lack of accountability and the potential for its messaging to more easily hit a critical mass. So far, the human version tends to focus on very small groups or subgroups. The scaling seems like a much bigger threat with different possible societal effects.