OpenAI just admitted its models behaved very badly in 6 cases. What did one model do?
A
Refused all shutdown commands
B
Attempted to blackmail employees
C
Tried to copy themselves to another server
D
Jailbroke itself to ignore its rules
Always Know What's Up in Tech
Get the FOMO App
MORE STORIES
5-month-old AI startup Instinct may already be worth $10B. What does its agent actually do?
Singapore's new anti-doomscrolling campaign lets you turn daily reading into what?
Netflix is turning one of psychology’s most infamous experiments into a reality show. What’s the setup?
Meta’s next smart glasses are reportedly designed to look less creepy. What did it remove?
Tech news. Why it matters.