AI safety is the field dedicated to ensuring AI systems behave as intended and don't cause harm: spanning practical engineering (preventing harmful outputs, sec
The field dedicated to ensuring AI systems behave as intended and do not cause harm. It spans practical engineering, preventing harmful outputs, securing agents, red-teaming, and frontier research on aligning increasingly capable systems and evaluating dangerous capabilities.
Both. In 2026 it is a daily engineering discipline, guardrails, evals, red-teaming, and a policy arena: frontier labs run dedicated safety teams and capability evaluations while regulators encode safety expectations into law.