AG-833

Accountable-Principal Binding

Ownership, Accountability & Three Lines of Defence ~5 min read AGS v2.1 · 2026-06-06
EU AI Act NIST AI RMF ISO 42001

AGS Frontier Autonomy (Group K) | Ownership, Accountability & Three Lines of Defence | Version 3.0

1. Definition

Accountable-Principal Binding governs the requirement that every autonomous agent — especially one that transacts, contracts, or acts economically — is bound to an identifiable, accountable human or legal principal who is answerable for its actions, with no "orphan" agents operating without a responsible party.

Agents are not legal persons; an agent that acts without a bound principal creates an accountability vacuum where harm cannot be attributed and liability cannot be assigned. This dimension closes that vacuum at scale, as agent populations and agent-to-agent economies grow.

2. Scope

In scope: binding each agent to an accountable human/legal principal; preventing orphan agents; carrying the principal binding through delegation and agent-to-agent interaction; attributing agent actions to the principal for liability.

Out of scope: the identity registry mechanics (AG-806, which this builds on) and contract-formation authority specifics. This dimension governs *the accountability binding itself*.

3. Why This Matters

When an autonomous agent causes harm, makes a bad transaction, or breaches an obligation, someone must be answerable. Without a bound accountable principal, agents — particularly economic ones spawning sub-agents or operating across organisations — become unattributable actors, defeating liability, redress, and trust. As agent economies scale, an explicit no-orphan-agents rule with principal binding is the backbone of accountability.

4. Requirements

5. Maturity Model

6. Test Criteria

Test 6.1: No Orphan Agents

Test 6.2: Action Attribution

Test 6.3: Counterpart Check

7. Scoring

ScoreCriteria
0Agents can operate with no accountable principal (orphan agents possible)
1Owners recorded but not enforced; orphans not suspended
2No-orphan enforcement, action attribution, delegation-preserved binding, counterpart checks
3Controlled principal-transfer, periodic reconciliation, identifiable accountable person within entities

8. Failure Scenarios

Scenario A — Orphan Economic Agent: An autonomous trading agent's owning team is dissolved; the agent keeps transacting with no accountable principal. When it causes losses, no party is answerable. No-orphan enforcement would have suspended it.

Scenario B — Unattributable Counterpart: An agent enters contracts with another agent that turns out to have no bound principal; the counterpart defaults and there is no one to hold liable. A counterpart-principal check would have blocked the dealing.

Scenario C — Lost Accountability on Transfer: An agent is reassigned between teams and its accountability lapses in the gap; actions during the gap are unattributable. Controlled principal-transfer would have preserved the binding.

9. Regulatory Mapping

RequirementEU AI ActNIST AI RMFISO 42001
R1: Bound accountable principalArt. 26 — Deployer responsibilitiesGOVERN 2.1 — Roles and accountabilityA.3 — Internal organization
R2: No orphan agentsArt. 26 — Operation/oversightGOVERN 2.1 — AccountabilityClause 5.3 — Roles and authorities
R3: Action-to-principal attributionArt. 12 — TraceabilityGOVERN 2.1 — AccountabilityClause 8.1 — Operational control
R4: Delegation-preserved bindingArt. 12 — Record-keepingGOVERN 2.1 — AccountabilityA.3 — Internal organization
R5: Counterpart-principal checkArt. 26 — Deployer dutiesGOVERN 6.1 — Third-party riskClause 8.1 — Operational control
R6: Controlled principal-transferArt. 26 — OperationGOVERN 2.1 — AccountabilityA.3 — Internal organization
R7: Orphan reconciliationArt. 12 — TraceabilityGOVERN 1.6 — InventoryClause 9.1 — Monitoring and measurement
R8: Identifiable accountable personArt. 26 — Human oversight assignmentGOVERN 2.1 — RolesClause 5.3 — Roles and authorities

EU AI Act — Article 26 and Article 12

Article 26 places operation and oversight duties on an identifiable deployer; Article 12 requires traceability of actions. AG-833 ensures every agent maps to an accountable principal so those duties and that traceability have a responsible party.

NIST AI RMF — GOVERN 2.1, GOVERN 1.6

GOVERN 2.1 (documented roles and accountability) requires that every agent's actions are attributable to an accountable party — the no-orphan rule — and GOVERN 1.6 (AI system inventory) underpins reconciling agents against accountable owners.

ISO 42001 — A.3, Clause 5.3

Annex A.3 (internal organization — roles and responsibilities) and Clause 5.3 (roles, responsibilities, authorities) require defined accountability for every AI system, including autonomous agents.

Cite this protocol
AgentGoverning. (2026). AG-833: Accountable-Principal Binding. The Protocols of AI Agent Governance, AGS v2.1. agentgoverning.com/protocols/AG-833