Positions as an end-to-end AI safety platform covering pre-launch testing, runtime guardrails, and ongoing red-teaming, backed by a proprietary adversarial data moat — but the differentiation claim rests entirely on scale metrics rather than outcome-based proof.