AI Safety.

The discipline that aims to prevent AI systems from causing unintended harm to users, operators, and society — covering technical, operational, and societal dimensions. It belongs to the AI & ML Security category of cybersecurity.

The discipline that aims to prevent AI systems from causing unintended harm to users, operators, and society — covering technical, operational, and societal dimensions.

AI safety is broader than traditional security: it addresses harms even when no adversary is present, such as accidents, bias, deception, runaway autonomous behaviour, dual-use misuse, and catastrophic or existential risk. Technical work includes alignment, interpretability, evaluation, robust training, monitoring, and capability elicitation. Operationally it covers responsible-scaling policies, model cards, deployment guardrails, and access controls. Institutions like the UK and US AI Safety Institutes, the EU AI Office, NIST (AI RMF) and frontier labs publish safety standards. AI safety is distinct from but overlaps deeply with AI security: insecure models often become unsafe, and unsafe models complicate security incident response.

Defences for AI Safety typically combine technical controls and operational practices, as detailed in the full definition above.

Common alternative names include: Frontier AI safety, Responsible AI.