Independent Submission A. Sogomonian Internet-Draft AIIF Intended status: Informational 9 June 2026 Expires: 9 December 2026 AIID: An Identifier Namespace for Autonomous Systems draft-sogomonian-aiid-namespace-00 Abstract Autonomous systems operate today without stable, globally unique identities. When an AI agent executes an action, there is no universally recognized identifier for that agent, no delegation record, and no accountability chain back to a human principal. This document defines the Autonomous System Identifier (AIID) namespace. An AIID is a structured, hierarchical identifier assigned to an autonomous system, execution environment, or agent operating within the Artificial Intelligence Internet Protocol (AIIP) architecture [I-D.sogomonian-aiip-architecture]. This document specifies the AIID syntax, hierarchical structure, registration model, operational states including safe mode, and revocation mechanisms. The governance model for the namespace, including selection of a root registry operator, is explicitly left open for community determination. Status of This Memo This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." This Internet-Draft will expire on 9 December 2026. Copyright Notice Copyright (c) 2026 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Table of Contents 1. Introduction 2. The Accountability Gap 3. AIID Syntax 4. Hierarchical Structure 5. Registration Model 6. Operational States and Safe Mode 7. Revocation 8. Namespace Governance 9. Security Considerations 10. IANA Considerations 11. Informative References 1. Introduction The AIIP architecture [I-D.sogomonian-aiip-architecture] defines a common execution model for autonomous systems based on a resolve-invoke-execute-receipt pattern. For this model to function across independently developed implementations, each participating system must be uniquely and stably identifiable. Existing identifier systems were not designed for autonomous execution contexts. Domain names identify information resources. IP addresses identify network endpoints. Neither provides the identifier structure required for accountable autonomous execution across organizational boundaries. This document defines the AIID namespace to address this gap. AIIDs are intended to be: * Globally unique * Hierarchically structured * Revocable * Bound to a verifiable root of trust * Subject to human operational control This document deliberately does not specify who operates the root of the namespace. Governance questions, including the selection of a root registry operator and the policy framework under which it operates, are addressed in Section 8 and are explicitly left open for community determination. 2. The Accountability Gap Today autonomous systems — AI agents, robots, automated services, and execution environments — operate without stable, globally unique identities. When an autonomous system executes an action today: * There is no universally recognized identifier for that system * There is no delegation record linking the action to a human principal * There is no verifiable receipt proving what occurred * There is no mechanism to trace responsibility back to an accountable human or organizational actor * Nobody is formally responsible for what the system did This situation is tolerable today because most autonomous actions are low-stakes and easily reversible. It will not be tolerable as autonomous systems increasingly operate in financial, medical, industrial, physical, and safety-critical environments. Consider what happens when an unidentified autonomous system: * Executes a financial transaction * Controls physical infrastructure * Makes a decision with legal consequences * Causes harm in a physical environment * Acts outside its intended boundaries Without identity, there is no accountability. Without accountability, there is no basis for governance, liability, or trust. AIID proposes to close this gap before it becomes a crisis. Every autonomous system that participates in AIIP execution carries a stable, traceable identity. Every action can be attributed. Every delegation chain traces back to a human principal who bears responsibility. Identity is therefore not a convenience feature of AIIP. It is the foundation on which accountability, delegation, policy enforcement, and human control are built. 3. AIID Syntax An AIID is a dot-separated hierarchical identifier of the following general form: Figure 1: AIID Syntax aiid ::= "aiid" "." organization "." identifier Examples: aiid.acme.robot-01 aiid.acme.agents.scheduler-7 aiid.example.agent-42 The components are defined as follows: aiid: The root label. Fixed as "aiid" for all identifiers in this namespace. organization: A label assigned to a registrant by the root registry. This label is unique within the root namespace. identifier: One or more labels assigned by the registrant to identify a specific autonomous system, agent, robot, or execution environment. Registrants may define sub-hierarchies within their assigned namespace. AIID labels consist of alphanumeric characters and hyphens. Labels may not begin or end with a hyphen. Labels are case-insensitive. The maximum total length of an AIID is 253 characters. 4. Hierarchical Structure The AIID namespace is organized as a hierarchy beneath a single root. Figure 2: AIID Namespace Hierarchy Root Registry | v aiid.* (root namespace) | +-- aiid.acme.* (organization namespace) | | | +-- aiid.acme.robot-01 (leaf identifier) | +-- aiid.acme.agents.* (sub-namespace) | +-- aiid.example.* (organization namespace) The root registry is responsible for: * Assigning organization-level namespace labels * Maintaining the root registry database * Publishing root trust material * Operating under the governance framework described in Section 8 Registrants are responsible for: * Managing identifiers within their assigned namespace * Maintaining accurate registration records * Ensuring AIIDs within their namespace are not misused * Managing operational states of agents under their namespace 5. Registration Model Registration of an organization-level AIID namespace requires: * Identification of the registrant (organization or individual) * Contact information * Intended use description * Acceptance of the applicable registration policy * Submission of a public key for namespace trust binding Registration is intended to be open to all autonomous system operators, vendors, and developers worldwide, with no restriction on the type of autonomous system, the implementing organization, or the geographic location of the registrant. Registration details, policy documents, and the root registry should be publicly accessible and machine-readable. 6. Operational States and Safe Mode Every AIID carries an operational state. The operational state determines whether an autonomous system may participate in AIIP execution flows. Human principals and registrants may set and change the operational state of any AIID within their authority. The defined operational states are: Active: The autonomous system is operating normally. It may participate in AIIP invocations, receive delegation grants, and produce execution receipts. Safe Mode: The autonomous system is operating in a restricted state. Safe mode is set by a human principal or authorized operator to limit the system's execution capabilities without suspending or revoking its identity. In safe mode an autonomous system: * Retains its AIID and registration * May be queried for status * May not accept new invocations unless explicitly permitted * May complete in-progress executions at the operator's discretion * Returns a safe-mode indicator in any execution response Safe mode is reversible. A human principal or authorized operator may restore an autonomous system to Active state at any time. Safe mode is the primary mechanism for human override of autonomous execution without permanent identity consequences. It allows operators to pause, investigate, and resume autonomous systems in a controlled manner. Suspended: The autonomous system is administratively offline. It may not participate in any AIIP execution flows. Its identity is preserved. Suspension is reversible. Revoked: The AIID has been permanently cancelled. The identifier may not be reactivated. See Section 7. Figure 3: Operational State Transitions Active <---------> Safe Mode | | v v Suspended <-------> Suspended | v Revoked (terminal) Human control over operational state is an architectural requirement of AIIP. The ability to place any autonomous system into safe mode at any time, from any point in the delegation chain, is a fundamental safety property of the AIID namespace. The mechanism by which operational state changes are communicated to execution environments is out of scope for this document and will be addressed in future AIIP specifications. 7. Revocation An AIID may be revoked by the root registry or by the registrant that issued it. Revocation renders the identifier permanently invalid for use in AIIP invocations and execution contexts. Revocation is distinct from Safe Mode and Suspension. Revocation is permanent and irreversible. It is appropriate when an autonomous system has been decommissioned, compromised, or has violated registration policy. A revocation record should include: * The revoked AIID * A timestamp of revocation * The revoking authority * A reason code (optional) Implementations should treat a revoked AIID as unauthorized. Execution environments should check revocation status prior to accepting an invocation. Revocation distribution mechanisms are out of scope for this document and will be addressed in future AIIP specifications. 8. Namespace Governance This document deliberately does not name a root registry operator. The author believes that the question of who operates and governs an identifier namespace for autonomous systems is at least as important as the technical design of the namespace itself, and that this question should be answered by the community rather than asserted by a specification author. Open governance questions include: * Should the root registry be operated by a single organization, a consortium, or a decentralized mechanism? * What accountability framework should govern the root registry operator? * How should registration policy be set and amended? * What fee structures, if any, are appropriate? * How should disputes over namespace labels be resolved? * What relationship, if any, should the AIID namespace have to existing identifier governance bodies? Possible governance models include operation by an existing Internet governance institution, creation of a new purpose- built body, a multi-stakeholder consortium of implementers, or a decentralized registry mechanism. Each model carries different tradeoffs in accountability, neutrality, cost, and operational reliability. The author's intent is that governance of the AIID namespace, if this work proceeds, should be determined through open community processes. 9. Security Considerations The AIID namespace is a trust anchor for the AIIP execution architecture. Threats to consider include: * Unauthorized registration of namespace labels * Impersonation of registered autonomous systems * Replay of revoked or suspended identifiers * Compromise of root registry trust material * Namespace squatting or abuse * Unauthorized modification of operational state * Circumvention of safe mode by autonomous systems The root registry should use hardware security modules or equivalent mechanisms to protect root signing keys. Root trust material should be published through multiple independent channels. Registrants should rotate signing keys periodically and revoke compromised keys immediately. Implementations must enforce safe mode and suspension states. An autonomous system must not be permitted to override or bypass its own operational state. Human override capability must be preserved at all times. 10. IANA Considerations This document has no IANA actions. Future revisions may define registries, namespace parameters, delegation formats, operational state codes, or discovery mechanisms if such actions are determined to be necessary. 11. Informative References [I-D.sogomonian-aiip-architecture] Sogomonian, A., "Architecture for the Artificial Intelligence Internet Protocol", Work in Progress, Internet-Draft, draft-sogomonian-aiip-architecture, March 2026, . [I-D.sogomonian-aiip-native-access-architecture] Sogomonian, A., "AIIP: Native Access Architecture for Autonomous Systems", Work in Progress, Internet-Draft, draft-sogomonian-aiip-native-access-architecture, June 2026, . Author's Address Aram Sogomonian (editor) Artificial Intelligence Internet Foundation (AIIF) Email: aiinternetfoundation@icloud.com