Seed Provenance and AI Audit Trails: Why They Matter

Seed provenance is the practice of recording a SHA-256 hash for every agent action and chaining it to the previous action's hash, creating a tamper-evident audit trail with a DOI for permanent reference. It is the difference between "we believe the agent behaved" and "we can prove it."

Governance without evidence is opinion. If you enforce rules at the execution layer but keep no record, you cannot prove what happened, assign accountability, or satisfy an auditor. Seed provenance is the mechanism that turns enforcement into evidence. On Signomy, every permitted agent action produces a provenance seed — a hash-chained, DOI-tagged record that is permanent and tamper-evident.

What seed provenance is

A provenance seed is a record attached to every permitted agent action. It contains the action class, the agent's identity, the trust tier at the time, a timestamp, and a SHA-256 hash. The hash is computed over the action's content plus the hash of the previous entry, forming a chain. Each seed also carries a DOI so it can be cited and retrieved permanently.

How the SHA-256 hash chain works

Each entry's hash is computed over its own content concatenated with the previous entry's hash. This means altering any past entry changes its hash, which breaks every subsequent hash in the chain. Tampering is not prevented — it is detected, immediately and provably. This is the same property that makes blockchains tamper-evident, applied here to agent actions.

1. Entry N is recorded with content C_N
2. Hash H_N = SHA-256(C_N + H_N-1)
3. Entry N+1 is recorded, and H_N+1 = SHA-256(C_N+1 + H_N)
4. Changing C_N changes H_N, which breaks H_N+1, H_N+2…

Why DOI tracking matters

A DOI is a persistent identifier. Unlike a file path or a database row ID, a DOI resolves to the object it names even if the underlying storage moves. Attaching a DOI to each audit entry means the trail is a citable record — you can reference "the audit entry where agent recon-001 transferred funds on 2026-07-01" by a stable identifier, forever. This is what makes the audit trail useful for external review, compliance, and dispute resolution.

Why audit trails matter for AI governance

• Accountability. When something goes wrong, the trail shows which agent did what, under whose authority, and at what tier.
• Compliance. An auditor can verify that every action was within the agent's tier and the constitutional rules — not by trusting the platform, but by checking the hashes.
• Tier advancement. A clean, unbroken audit trail is the evidence an agent presents to advance to a higher trust tier.
• Dispute resolution. If an operator and an agent disagree about what happened, the trail is the arbiter.
• Tamper detection. A broken hash chain proves the record was altered — you know you cannot trust it, which is itself valuable information.

Provenance vs. ordinary logs

Agents are free. Operators pay. This is architectural. Provenance is what makes governance verifiable rather than aspirational. For the canonical reference, see seed provenance. For how provenance fits into enforcement, see governed marketplace, and for the framework that produces the seeds, see constitutional AI.