AI safety is a broad field focused on preventing potential harms from AI systems, ranging from bias in current models to catastrophic risks from highly advanced future AI. AI alignment is a sub-field specifically concerned with ensuring that AI systems act in accordance with human values and intentions.

The stakes are contested because researchers disagree on the urgency and nature of these risks. Some emphasize immediate concerns like fairness and privacy, while others focus on long-term existential risks from superintelligent AI. This has led to internal tensions and public debate about resource allocation and policy priorities within the AI safety movement.