The Poisoned PII Problem

The 2017 Equifax breach affected about 147 million people. Events at that scale made a basic security mistake harder to ignore: a fact can identify someone without remaining secret enough to authenticate them.

Names, addresses, dates of birth, and government identifiers still matter for identity resolution and regulated customer programs. Their value as shared secrets is weaker. A system should not treat knowledge of a recoverable biographical fact as strong proof of current control or intent.

Current Guidance Has Moved Past KBA

Knowledge-based authentication asks a claimant to answer questions derived from personal history or records. NIST Special Publication 800-63A-4 says knowledge-based authentication and knowledge-based verification must not be used for identity verification. NIST guidance also excludes KBA as an acceptable authenticator.

That does not make every use of personal data invalid. It separates 2 jobs. PII can help match a claimant to a record. Authentication needs stronger evidence of control, and authorization needs evidence that the authenticated actor may perform the specific action.

An Agent Needs a Different Evidence Chain

Software agents do not possess a human biography. They operate through credentials, workloads, wallets, tools, and mandates issued by a person or organization. Presenting the principal's PII does not establish which agent is acting or whether that action fits the mandate.

A useful agent identity record binds the agent to an accountable principal and a lifecycle state. A transaction decision then checks the current scope, wallet or account, counterparty, amount, runtime signals, and any required tool version. Identity is durable context. Authority is evaluated for the action at hand.

What a Signature Proves

A valid signature identifies the signing key and shows whether the signed bytes changed. Its meaning depends on the issuer, the claims, the evidence behind those claims, and the relying party's policy.

A signature alone does not prove that the underlying claim is true, that a mandate covers the current action, that a runtime is uncompromised, or that an allowed action later succeeded. Those facts need their own evidence and, when applicable, separately signed records.

A Practical Control Stack

• Resolve the principal. Use appropriate identity-proofing and compliance evidence for the person or organization behind the agent.
• Authenticate the agent. Verify a credential or key bound to the agent and check its issuer, status, and expiry.
• Evaluate authority. Compare the current action with the principal's mandate, limits, counterparty, and time window.
• Preserve the decision. Record the evidence, unavailable checks, rules version, verdict, and issuer signature.
• Record later events separately. Execution, settlement, and fulfillment remain distinct facts after a verification decision.

Related Academy Briefs

• Know Your Agent
• A Signed Decision Is Not an Execution Outcome