{"page":"/blog/ai-agents-data-leakage","modified":"2026-11-10","license":"https://creativecommons.org/licenses/by/4.0/","columns":["Field","Purpose"],"rows":[["model_location, provider_terms","Records which column of the decision table applies, and the contract that makes it acceptable"],["data_classes","One verdict per data class, matching the decision table"],["clean_room_outputs.min_users_per_row","Mirrors a 50-user aggregation threshold such as Ads Data Hub's; adjust to your clean room"],["max_queries_per_hour","Stops an agent spending a privacy budget built for human query rates"],["tools.outbound","Empty on purpose: this agent reads data, so it gets no tool that sends data out"],["network_egress_allowlist","Enforced by the runtime's network layer, not by the model"],["logging.redact_fields","Identifiers removed before any log is written"],["human_approval","Exports and handoffs to other agents wait for a named role"]],"sourceUrls":[[],[],[],[],[],[],[],[]]}