One website was enough to hijack Microsoft’s AI agent
Microsoft’s security researchers have uncovered a worrying attack technique that shows how AI agents can become a security risk when they are given too much trust. The attack, nicknamed AutoJack, targeted an early development...
