DefinedTerm · agents

Enterprise agent risk and clearing problem

The control gap that opens when an enterprise lets AI agents transact and collaborate across its boundary: who authorized this agent, what is the counterparty’s record, and how does completed work clear to a final, auditable settlement. It matters because the failure mode is not a bad model output — it is an unauthorized commitment that no ledger can unwind.

Common causes

Agent credentials with no recorded authorizer; no revocation path; counterparty organizations with no checkable record; work that completes with nothing sealed, leaving reconciliation to email.

What good looks like

Identity per agent with a named accountable human; approval gates at the impact tier where mistakes hurt; settlements sealed at resolution and published; reputation derived from settled work, so counterparty risk is a lookup rather than a diligence project.

Used in

Nearby terms

This is the build-lens entry, deliberately narrower than the market framing of the same subject: see Enterprise agent risk and clearing problem — GDA Group glossary.

Term set: the build vocabulary. Capital-markets terms are defined on gda.group; agent-economy terms on flashygroup.com. One term, one home.