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

I wonder what happens when bots find other suspects are bots ... deploy more bots? More security?

Not unlike the situation where undercover cops ended up surveilling other undercover cops...

https://www.theguardian.com/uk-news/2020/oct/28/secrets-and-...

Could be a lot of fun creating and deploying adversarial bots to lure these things down rabbit holes.

Mr. Anderson vs. Mr Smith and Mr. Smith and Mr. Smith, et. al.

https://youtu.be/poLPKdc-zTY?si=eRYs8-z0jN2JLsjq

There's less than zero incentive to prevent any of this because both agencies can use the make-work to justify their budgets and the vendors make out like bandits.

And we all pay for it.

Do you want to explain to a judge that "it was just AI which made credible claims of commiting crimes"?

Not my definition of fun.

I don't disagree it would be unwise.

But as a thought experiment it is very interesting.

From outside the jurisdiction? Note how hard it is for the US to extradite global financial scammers, for example.

"What does this GPU cluster do?" gestures at rack "Ah, it torments authoritarian AI vendors."

>From outside the jurisdiction?

How sure are you that they do not cooperate with your local law enforcement already?

Then we end up in a real-world The Man Who Was Thursday…

(Spoilers) The anarchist association has seven members, eventually the members discover there’s only one real anarchist and six policemen.

The number of countries friendly with the US diminishes by the day.

(US citizen, have helped with standing up non profit infra outside the US to avoid US reach, both legally and technically)

Sure. I definitely think it would be interesting to see, nevertheless there is a risk of severe negative consequences.

Why would you risk your freedom over this.

And the number of countries who would take a terrorism warning from the US seriously has not diminished at all. You are forgetting that it is a crime in your jurisdiction as well, so it will be pursued by your government. No extradition necessary to ruin your life.

You've taken something from a fun thought exercise to "why are you executing on this!?" We're just having fun, but I encourage anyone to freely take the idea and run with it. The world is a big place.

Note the concern you share, which shows how far we've slid towards said authoritarianism. If you're afraid to even talk about it, the damage done is evident. Fear is a feeling, danger is objective.

But this fun thought exercise is also commiting a crime. That should be pointed out.

>Note the concern you share, which shows how far we've slid towards said authoritarianism.

Authoritarianism? This is about people saying they will commit acts of terrorism.

I absolutely think people should be critical of this type of police work, but if someone makes credible threats of commiting acts of terrorism they should obviously be taken in by police.

You are mistaken. Talking about it is not a crime, and if you believe having an AI honeypot another AI is, please provide legal citations or a legal opinion. This is no different than generating adversarial text programmatically to trap a bot, as there is no intent. No one is suggesting terrorism, only poisoning chatbots. Of course, if honeypot chatbots are outlawed, only outlaws will have them. Interesting times ahead, for sure.

With bots it seems even more automatic.

Whole system could be just billed on "usage".

Again. You will be the precedent. It will be your case which decides whether this is a crime or not.

But if you believe the prosecution won't try you have to be deluded.

They'll never go for it because the plausibility of their "we're doing something" pretext is what justifies the entire racket. Paying them to go away wouldn't accomplish that goal.

I read this entire thread before noticing how oddly relevant your respective usernames were to the conversation

Protesting is not terrorism. Well, not in most countries.

Sounds like Woodie Guthrie's guitar