Scores AI interactions for prompt injection, PII, harmful requests, and refusals.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.