Safety, Ethics & Governance · beginner · concept 175 of 187
AI Safety
AI safety is the umbrella field concerned with preventing AI systems from causing harm, covering accidents, misuse, robustness failures, evaluation, security, and governance. Alignment, getting a system to pursue the objectives its developers intended, is one subfield inside it, not a synonym. AI ethics overlaps but asks normative questions about fairness and consent rather than engineering ones. Practitioners often treat near-term harms and long-term catastrophic risk as rival agendas; most working labs, standards bodies like NIST, and national institutes fund both, and which deserves priority is contested.
Key terms
Learn these first
Where you meet it in the real world
Frontier-model release safety cases, NIST AI RMF in enterprise risk programs, national AI Security Institute evaluations, jailbreak and misuse defenses in deployed chatbots
Videos
Computerphile · YouTube
Guides and articles
Courses, papers, and more