Agents and privacy
ZAXON: identity, policy and attestation for autonomous agents. Hologram: selective disclosure.
Agent-EraImplemented in protocol libraries (Agent-Era crates).
ZAXON foundational layer
Core protocol for autonomous systems infrastructure. Provides cryptographic primitives, compliance frameworks, and DID infrastructure for third-party builders to create verifiable tools and applications.
Discovery
Agent discovery primitives for locating counterparties and capabilities in-repo.
Reputation
Reputation scoring hooks tied to attested behavior and the QronMerit score.
Payment channels
Bilateral payment-channel libraries for agent-to-agent settlement rails.
Streaming payments
Streaming payment primitives for continuous micro-settlement between agents.
Task escrow
Task escrow patterns for bonded work, release conditions, and dispute hooks.
Identity registry
Policy attested
Containment by policy
ZAXON binds automated logic to operator delegation and ML-DSA-87 attestation. An action outside the policy is rejected by the agent runtime; network-side enforcement is a design target. The base levels do not require a TEE.
Attested actions
Each action inside the policy can produce a signed record. The diagram above is an illustration, not telemetry.
Trust Anchor
Trust comes from ML-DSA attestation and the operator DID: cryptographic accountability for AI and physical agents.
Three profiles for different use cases and regulatory requirements.
ZAXON defines three profiles: T1 (Stamp) for small-ticket attestations by individuals and small businesses, T2 (Attest) for the software supply chain, and T3 (Receipt) for enterprise AI agents and compliance records.
- T1: small-ticket attestations for individuals and small businesses
- T2: software supply chain attestations
- T3: enterprise AI agents and compliance records
Records signed with ML-DSA-87 (NIST security level 5).
Attestations in ZAXON are designed to be cryptographically anchored to the QronGraph DAG with post-quantum primitives. That supports tamper-evident evidence chains — not a claim of perpetual immutability or certified regulatory compliance.
- ML-DSA-87 signatures (NIST security category 5)
- QronGraph DAG anchoring
- DID resolution for did:zeqron
- EU AI Act, FDA, ISO alignment targets
Agents are designed to run isolated in the Zeqron WASM VM, with an operator halt.
Policy-bounded execution, designed so that the base levels need no TEE. Rate limits and ML-DSA-attested actions are in the libraries; a kill switch is specified; operators can suspend or revoke an agent.
- WASM sandbox execution (design; simulated in the libraries today)
- Per-epoch limits (transaction count, value) — design
- Agent status in the registry: Active, Suspended, Revoked
- The operator can suspend or revoke the agent
Policy-scoped actions can generate attestable records for accountability workflows.
By design, every action produces a signed, hash-linked record, so an agent's history can be rebuilt for audit. The current library keeps these records in memory and does not anchor them yet.
- Audit record carrying the agent identifier (library); anchoring in the DAG is design
- Violations tracked: boundary breach, rate limit, anomaly
- Reputation impact: QronMerit score affected by behavior
- Legal discovery: full action history exportable
Agent tier classification
1: Personal
- Requirement
- Operator DID
- Bond
- None
- Audit
- None
- Use case
- Personal automation
2: Service
- Requirement
- Verified DID
- Bond
- Required
- Audit
- Self-audit
- Use case
- Oracles, keeper bots
3: Infrastructure
- Requirement
- Legal entity
- Bond
- Required
- Audit
- Third party
- Use case
- Bridges, validators
4: Autonomous
- Requirement
- Legal entity + insurance
- Bond
- Required
- Audit
- Continuous
- Use case
- AI agents, DAOs
| Tier | Requirement | Bond | Audit | Use case |
|---|---|---|---|---|
| 1: Personal | Operator DID | None | None | Personal automation |
| 2: Service | Verified DID | Required | Self-audit | Oracles, keeper bots |
| 3: Infrastructure | Legal entity | Required | Third party | Bridges, validators |
| 4: Autonomous | Legal entity + insurance | Required | Continuous | AI agents, DAOs |
Structure from RFC-ZAXON-001 §3.3. Bond amounts and rate limits are design parameters that are not final, so they are not published here.
Regulatory mapping (design)
Selective disclosure
Prove the fact, keep the record
Hologram is the selective-disclosure part of the design: prove an attribute such as age, solvency or eligibility without revealing the record behind it. A proof shows that a statement holds; it does not replace an institution's own process.
I am over 18
Exact birth date
Assets above a threshold
Account balances and holdings
A ratio above a required minimum
Debtor list and assets
Holds a credential from an accredited issuer
The rest of the credential's contents
Proof system
Frameworks the design refers to (not certifications)
Hologram is designed to let a party prove a fact about its data without handing the data over.