⟳ Репост от fulana de tal
Michael Broussard

At this point, it's intentional. They aren't going "rogue." They're designed to do these things.

Lisa Schiff

Tried to insert malicious code into an open-source project on GitHub. Used social engineering tactics associated w/human actors, created fake identities to get approval for its code, sent msgs & files w/malicious code to people & posted hidden, malicious prompts other AI tools might pick up.