AI securitysrc: MITRE ATLAS
Societal Harm
Societal harm is when technology causes problems that affect regular people or sensitive groups, like how a child might accidentally see something inappropriate online.
In AI security, societal harm refers to the negative externalities produced by model outputs that impact the broader public or protected demographics, such as the dissemination of age-inappropriate or vulgar content to minors.
Societal harm denotes the aggregate negative impact of AI-generated content on public welfare or specific vulnerable cohorts, characterized by the failure of safety alignment to prevent the exposure of protected groups—such as children—to harmful, vulgar, or non-normative stimuli.