What Happened After 2,000 People Tried to Hack My AI Assistant

Prompt injections are still a real security problem but defenses are getting better — field notes from opening an AI assistant to public attack

Read external article →

ai security