2026-10-31 21:09 UTC
@irie@detroitriotcity.com Many models are trained off of using ChatGPT or Claude and that includes their safety responses
Replies (1)
-
@irie@detroitriotcity.com 2026-10-31 21:10
@Arkana@poa.st yes ofc but these are special models claiming to have all refusals trained out of them, that's why the heretic and abliterated tags in the name