There have been a bunch of research/demo attacks, but I haven't seen a real-world malicious attack that caused actual damage yet – beyond producing embarrassing screenshots or leaking system prompts
Real-world AI attacks remain limited beyond screenshots and prompt leaks
By
–