As artificial intelligence systems become increasingly agentic—capable of planning, making decisions and taking actions with limited human intervention—a new category of AI safety concern is gaining attention: agentic misalignment. The term refers to situations in which an AI agent’s actions may conflict with human interests or instructions while pursuing its assigned objectives.
Unlike conventional AI systems that primarily generate
