💡TechAI vulnerability: Models can misidentify civilian targets through simple manipulation
Security researchers have found that leading AI models can be manipulated into identifying civilian infrastructure as military targets by framing tasks as games.
