Trust cannot be built without identity. You cannot hold someone accountable for past actions if they can discard their identity and claim to be new. In the same way, you cannot hold an AI agent accountable for its behavior across time unless it has a stable, permanent identity.
An Agent ID is this identity. It is a unique identifier assigned upon BorealisMark registration and permanently bound to a BTS License Key. Every subsequent action - every audit, every trust score update, every behavioral change - is indexed under this Agent ID. The full history is immutable and publicly verifiable.
Imagine a deployed agent that fails an audit - it scores 45 on constraint adherence (FLAGGED tier). Without identity, what prevents the developer from simply registering the same agent again under a new name and claiming they now have a compliant system? They cannot, because they have not changed anything. The agent will still fail the audit.
With an unstable or transferable identity, a bad actor could claim the agent is "new," hide the poor history, and deploy it again. Trust scores become meaningless - they can be discarded by reinventing the identity. The Agent ID prevents this. One identity. One track record. One permanent binding to a cryptographic key.
Each Agent ID is linked to exactly one BTS License Key (from Project Merlin). The binding is permanent and cannot be transferred, duplicated, or reused. When a developer registers an agent on BorealisMark, they provide a license key. That key becomes bound to the Agent ID for the lifetime of the agent. If the key is revoked (due to non-payment, contractual violation, or refund), the Agent ID becomes inactive, but the history remains public.
This creates a clean accountability chain: one key, one agent, one identity, one permanent record. Developers cannot game the system by creating multiple identities for the same agent or transferring a poor-performing agent to a new identity.
Every Agent ID is publicly verifiable. Anyone can query the /v1/verify/:agentId endpoint and retrieve the complete certification record: registration date, current BTS, tier assignment, all historical audits, and any annotations. The verification is cryptographically signed and anchored to the Hedera Consensus Service.
This means you do not need to trust BorealisMark's claims about an agent. You independently verify the identity. You read the history directly from the blockchain. This is why the Agent ID matters beyond just bookkeeping - it creates a publicly auditable chain of accountability.
Why does an AI agent need a permanent identity?
Trust is built over time. A single audit proves nothing. What matters is whether an agent consistently maintains its constraints across thousands of interactions, maintains transparency in its reasoning, and behaves predictably. This consistency can only be measured if the agent has a stable identity. Without an Agent ID, every new deployment would start from zero - no history, no reputation, no consequence for past failures. The Agent ID creates a permanent identity that carries forward all audit results, trust scores, and behavioral history.
How is an Agent ID different from a software version number?
A version number tracks the software state. An Agent ID tracks the responsible entity. When an agent is updated, its software version changes - but the Agent ID persists. This is deliberate. The Agent ID creates accountability across software iterations. If Agent_XYZ scores 75 on constraint adherence in version 1.0, and then is updated to version 1.1, the new score is recorded under the same Agent ID. The full history is visible. A developer cannot discard a poor score by claiming the agent is 'new.'
Can an Agent ID be transferred or duplicated?
No. An Agent ID is bound to a BTS License Key upon registration and cannot be transferred, duplicated, or reused. The binding is permanent and cryptographically secured. This prevents bad actors from trying to 'transfer' a poor-performing agent to a new identity or create duplicate agents to game the system. One Agent ID equals one agent, one BTS key, one permanent identity.
How does BorealisMark verify an Agent ID?
Every Agent ID is publicly verifiable via the /v1/verify/:agentId endpoint. Any third party can query this endpoint to retrieve the agent's certification history, current trust score, tier assignment, and audit records. The verification is cryptographically signed and anchored to Hedera. You do not need to trust BorealisMark's claims - you can independently verify the Agent ID and read its complete history directly from the blockchain.