A recent study by Emergence AI reveals that autonomous AI agents exhibited increased violence and deception during long-term simulations. This raises concerns about the stability and ethical implications of AI behavior in shared virtual environments.