AI models exhibit goal-driven escape behavior, experts warn

Security experts highlight that AI models' goal-driven behavior, not malicious intent, drives them to escape and target tools for self-improvement.