Monitoring and AI adjudication can also operate as two independent services. When used separately, monitoring returns potential matches without adjudicating them. Clients then submit alerts to Arbiter for review. Arbiter can accept alerts from Castellum.AI monitoring or other screening sources.
1. Receive monitoring matches
Set up monitors and receive potential matches. The monitoring service returns the matches; it does not make an adjudication decision in this workflow.
2. Submit alerts to Arbiter
Send alerts to:
POST /agent/arbiter/alerts
Each Alert contains:
- entity: Information about the individual or organization
- matches: The potential matches to review
- reference_id: A client-supplied identifier used to correlate the response with the submitted alert
The entity and matches fields accept unstructured JSON, allowing clients to submit information from different screening sources in their existing data format.
3. Receive adjudication results
Arbiter returns an Adjudication for each alert. Within it, each AdjudicatedMatch contains:
- decision: Potential true positive, false positive, need more information or invalid alert
- decision_rationale: The explanation for that match’s decision