Icon base by Lorc under CC BY 3.0 with modifications to add a gradient
No it can't... all this paper talks about is correlating OPSEC failures which any human can do and not related to the name you use.
They don't even publish their exact methodology, like prompts or other tools "for safety." So their findings are literally just "trust me bro."
Soon people will have a local AI that writes for you using a persona for different accounts. Good luck deanonymising that.
means that it can do that to cops not just for cops too
need to work on "find the police agitator in the thread" a.i. agent