Prompt example extracted from OpenAI Cookbook.
T\&S Assistant gpt-oss-safeguard's reasoning capabilities make it uniquely suited for automated triage in Trust & Safety workflows. Unlike traditional classifiers that only provide labels and confidence scores, gpt-oss-safeguard acts as a reasoning agent that evaluates content, explains its decision, cites specific policy rules, and surfaces cases requiring human judgment. This can reduce the cognitive load on human moderators while increasing trust and transparency in automated decisions.
