Post #3970554

Look I get that hugging face is a 1st hand source, but they really didn’t provide any evidence or rationale explaining how they know the attack was LLM-based and they have a lot to gain by not only naming a Chinese open-source LLM but by also feeding into fear-based marketing… for all I know the models they used were hallucinating the “breach.” Nothing tampered with, no exfil, what even fucking happened lmao. I’m not buying this. Yet.