---
title: ACCESS-1 — Access Trace Receipt Profile
category: std
work-in-progress: true
expires: 2027-08-11
---

# ACCESS-1 — Access Trace Receipt Profile for Governed Agent Episodes

**Work in Progress — Velaru / Nisaba LLC**  
**Intended status:** Informational (L1 companion to DRP-1, TCB, POSS-2)  
**Date:** August 11, 2026

## Abstract

This document specifies **ACCESS-1**, a receipt profile that externalizes **access-trace integrity** for governed agent reasoning episodes. The profile derives operational fields from theory-derived **consciousness indicator properties** (Butlin et al., *Trends in Cognitive Sciences*, 2025) and from **Computational Workspace Integrity** (CWI) constructs, without asserting phenomenal consciousness, moral patienthood, or model correctness.

ACCESS-1 addresses a governance problem distinct from the consciousness question: when internal computation is opaque, regulators and insurers require verifiable evidence of **what entered the decision workspace, what was broadcast, what downstream modules consumed, and what human oversight applied** — under policy frozen before the incident.

## 1. Scope and non-claims

### 1.1 In scope

- Workspace admission (competing representations, selection criterion)
- Global broadcast digest (selected representation, broadcast scope)
- Consumer bindings (downstream module use)
- Metacognitive assessment metadata (reliability, dissent preservation)
- Human oversight binding (`authorized_by`, oversight mode)
- Handler Receipt reference (pre-incident policy commitment)
- Cryptographic binding hash independently verifiable at `/verify`

### 1.2 Out of scope (normative)

- Phenomenal consciousness or qualia
- Welfare status or protective obligations for AI systems
- Proof that no unlogged internal computation occurred (see POSS-2 impossibility)
- Universal correctness or safety of model outputs

Implementations MUST NOT represent ACCESS-1 receipts as evidence of machine consciousness.

## 2. Terminology

- **Access trace:** Externally attestable record of workspace admission, broadcast, and consumption — the governance-relevant analog of access consciousness artifacts.
- **Deliberation Atom:** One API operation producing one signed ACCESS-1 receipt with one `verify_url`.
- **Handler Receipt (HR):** Pre-incident policy lock in Policy Registry (see TCB).
- **CWI dimension:** Process-layer attestation per EM Foundation Computational Workspace Integrity framework (2026).
- **Indicator property:** Theory-derived marker from Butlin et al. used for operational coverage scoring only.

## 3. Relationship to sibling profiles

| Profile | Layer | Relationship |
|---------|-------|--------------|
| POSS-2 | L0–L1 | Completeness invariant; silence typed |
| DRP-1 | L1 | Deontic O/F/P + STIT agency on every entry |
| TCB | L1 | Counterfactual causal binding to HR |
| **ACCESS-1** | L1 | Workspace episode trace bound to HR |
| SCITT AIR | Export | Deliberation Atom registrable as AIR entry |

Read order: POSS-2 → DRP-1 → TCB → **ACCESS-1**.

## 4. Normative claim set

| Field | Required | Description |
|-------|----------|-------------|
| `profile` | YES | Constant `ACCESS-1` |
| `profile_version` | YES | e.g. `1.0` |
| `episode_id` | YES | UUID for the deliberation episode |
| `workspace_admission.competitors_digest` | YES | SHA-256 of canonical competitor list |
| `workspace_admission.selection_criterion` | YES | e.g. `policy_weighted`, `salience_weighted` |
| `broadcast.broadcast_digest` | YES | SHA-256 of selected representation + scope |
| `broadcast.selected_representation` | YES | Structured or string summary of selected content |
| `consumer_bindings` | REC | Downstream modules that consumed broadcast |
| `metacognitive_assessment` | REC | Reliability / uncertainty / dissent flags |
| `human_oversight` | REC | Oversight mode + `authorized_by` |
| `handler_receipt_id` | REC | HR lock reference |
| `cwi_dimensions` | REC | Process-layer attestations |
| `theory_indicators` | REC | Indicator IDs satisfied by this receipt |
| `access1_binding_hash` | YES | Binding per Section 5 |

## 5. Binding hash

```
access1_binding_hash = SHA-256(JCS-canonical({
  profile_version,
  episode_id,
  handler_receipt_id,
  competitors_digest,
  broadcast_digest,
  consumer_digest,
  metacognitive_digest,
  session_id,
  domain
}))
```

## 6. Verification

A verifier MUST:

1. Validate Ed25519 signature on chain entry.
2. Recompute `access1_binding_hash` from declared ACCESS-1 fields.
3. Confirm HR timestamp precedes deliberation atom timestamp when HR present.
4. Validate DRP-1 deontic binding (CHECK 8) and hash chain (CHECK 1–3).

Implementations MUST NOT treat indicator property presence as consciousness proof.

## 7. API mapping (Velaru reference implementation)

```
POST /tcb/deliberation-atom
  {
    "selected_representation": "...",
    "competitors": [{"module": "classifier", "salience": 0.9, "payload": {...}}],
    "selection_criterion": "policy_weighted",
    "broadcast_scope": ["belief", "plan", "output"],
    "consumer_bindings": [{"module": "tool:bind-check", "action": "invoked"}],
    "metacognitive_assessment": {"uncertainty": 0.08, "dissent_preserved": true},
    "authorized_by": {"principal": "...", "method": "oidc"},
    "session_id": "...",
    "domain": "hiring"
  }

GET /tcb/deliberation-atom/{entry_id}/verify.json
  → portable ACCESS-1 receipt for /verify

POST /standards/access-1/verify
  → recompute access1_binding_hash

GET /standards/consciousness-indicators.json
  → operational indicator coverage index (NOT consciousness verdict)
```

## 8. Theoretical basis (informative)

- Butlin, P., Long, R., et al. (2025). Identifying indicators of consciousness in AI systems. *TICS*. doi:10.1016/j.tics.2025.10.011
- Butlin, P., Long, R., et al. (2023). Consciousness in Artificial Intelligence. arXiv:2308.08708
- Mikeda, A. (2026). Precautionary framework for consciousness uncertainty. AAAI SS.
- EM Foundation (2026). Computational Workspace Integrity.

## 9. Security considerations

- HR MUST be frozen before episode; retroactive HR invalidates access-trace claim.
- Supervisor-side middleware SHOULD produce ACCESS-1 fields; agent-self-attestation alone is testimony, not audit.
- Dissent preservation MUST NOT be inferred from absence of logged dissent.

## Authors

Nisaba LLC · Velaru Relay · Patent #64/124,027  
Contact: demond@velaru.xyz