> You could also make the same argument about check forgery or wire fraud.

You couldn't, because unlike all the things we discussed about AI, those are actually harming the victim in real, direct terms.

Having some model who doesn't know or care about you talking just like you, without claiming it's actually you, to a bunch of people who don't know or care you exist? That's zero actual damage to you.

> "You're not special" is a bad argument for why someone impersonating you isn't a problem. I mean even Black Mirror explained why impersonation is a problem in the very first episode many years ago.

I remember that episode. However, my argument isn't "you're not special, therefore impersonating you isn't a problem" - it's "you're not special, so what you think is impersonating you probably isn't, and even if, it doesn't hurt you in real terms", combined with "anything that's actual impersonation and/or hurts you directly was already possible, and AI currently doesn't impact this at all". Someone wants to screw with you? You're being targeted. AI might make the attacker's job a bit easier, but it's still someone going through the effort, vs. a background process on the Internet everyone seems to think LLMs are.

Also: there's an inverse relationship between weight of accusations and social proximity. A specific person you know (and other people know you know) accusing you of something? It's a problem. Some random comments from random accounts, accusing you and 100 different people of something? Most people won't believe it.

(Except when it's about child abuse. People are extremely sensitive to this - just bringing up the term and a name in the same sentence can ruin the victim's life.)