AI safety
Definition: AI safety is the field aiming to make AI systems reliable, controllable and beneficial, while limiting their risks and harmful uses.
It spans alignment, risk evaluation, robustness and governance. Anthropic has made it its central mission, with a research-driven approach.