Standards

Open by construction. The formats our verdicts ride on are published open specifications anyone can read, implement, or object to — so the proof you rely on is never locked inside one vendor.

Each is a versioned RFC-style document with a public comment window, a GitHub repository for proposals and objections, and a commitment to transfer editorial control to a neutral body once external adoption and standards-track recognition justify it. None is a paid product.

PIP Whitepaper
v1.0
A Policy Information Point for the agentic internet
v1.0 — Published 2026-07-28 · published 2026-07-28

Architecture background for the specs on this page: the PEP/PDP/PIP model, the four obligations a PIP owes a PDP (facts with provenance, coverage honesty, a freshness contract, verification without trust), and how the published surfaces implement that contract.

Signed Attribute Bundle
v1.0 (draft)
The output contract of a Policy Information Point
v1.0 — Public comment through 2026-10-24 · published 2026-07-26

What a PIP owes a PDP: per-source facts each labeled with a coverage tri-state (observed_clean / never_observed / degraded — an observation, an honest blind spot, or a failure, never a guess), a freshness contract (valid_until, stale_if_error), and an Ed25519 signature binding via key_id. Facts, not verdicts — the trust decision stays in your policy engine. Live on every POST /v1/verify response; includes an OPA/Rego consumption example and a streaming (NDJSON) form with a published conformance vector.

Reconciliation Verdict
v1.0
Claim-vs-conduct verdict for any cryptographic key
v1.0 — Live · published 2026-06-16

The reconciliation layer above all roots of trust. One verdict, keyed on an Ed25519 key, that reconciles what the key can prove about itself (its attestation tier across roots — a bare self-attestation or a RATS/EAT hardware token) against what the network has seen its subject do (Scry × Sigil × GhostRoute × Tracker). Surfaces over-claim contradictions no single root can. Ships as a self-verifying receipt with an RFC 6962 inclusion proof and witness cosignatures — verifiable offline, no call back to TunnelMind.

OAI
v1.0
Observed Actor Identifier
v1.0 — Public comment through 2026-08-12 · published 2026-05-14

Open identifier standard for entities that observe, profile, or act against users, devices, and networks. Free resolution, permanent canonical identifiers, signed observations. CVE-style editorial model.

Signed Observation
v1.0.0
Ed25519-signed sensor observation wire contract
v1.0.0 — Frozen 2026-05-30 (ADR-A3) · published 2026-05-30

The wire shape every TunnelMind sensor speaks when submitting a first-party observation to the corpus. Compact-JSON canonicalization, Ed25519 signature over the record, producer-agnostic (Familiar today; commodity sensors, contribute-and-earn third parties, and future microkernel devices in future). Single canonical schema URL pins forward compatibility.

Compliance Ledger
v1
Tamper-evident, customer-configurable agent-action record
v1 — Live · published 2026-06-13

A per-customer, append-only, hash-chained ledger of an agent’s verdict receipts, with Ed25519 signed checkpoints and regime-mapped signed exports (EU AI Act Art.12, DORA, NYDFS Part 500, HIPAA, PCI DSS, SOC 2) in signed_json / CSV / EAT / STIX. You pick the regime, retention, and format; an auditor verifies the chain independently against the published receipt key.