Getting Started with HAIEC

Create or select an AI System in HAIEC. HAIEC organizes evidence and assurance around a bounded AI System, so every evaluation, action, and receipt is scoped to a specific system.

Register and connect the supported evidence sources relevant to that system. Examples include source repositories, application access, provider/IAM evidence, and runtime evidence where connected. Registered does not mean connected. Connected does not mean evaluated.

Run or attach supported evidence producers: static analysis, runtime evidence where connected, policy, provider/IAM evidence, or self-reported/uploaded evidence where supported. Missing evidence remains missing — it does not silently become PASS.

Review what the AI system can reach: identities, application access, credential evidence, actions, APIs/tools, potential consequences, policy context, and unknown relationships. HAIEC does not promise universal discovery — the map shows what available evidence establishes.

Use Evaluation Runs to coordinate evidence collection and evaluation across supported producers. An evaluation run produces operational evidence and scores. An evaluation run is not an Assurance decision.

Review the canonical Assurance outcome: ALLOW, REVIEW, or BLOCK. ALLOW means the system is permitted within the evaluated scope and available evidence — it is not a universal safety or compliance claim. UNKNOWN, NOT_ASSESSED, PARTIAL, FAILED, TIMEOUT, UNSUPPORTED, and NOT_RUN never silently become PASS.

Review the Decision Receipt and public verification where supported. Verification confirms bounded record integrity and the recorded decision. It does not certify compliance, regulatory approval, or universal safety.