Security researchers have identified a technique that could allow attackers to execute malicious code within AI-assisted ...