VAPVERIFIABLE AI PROVENANCE

30 October follow-up: evidence preservation, rechecks and guide notices

VAP #33 binds publication, disclosure and authorization; #34 covers scoped erasure; #35 preserves 53 original evidence cases and an offline kit; #36 generates bilingual reports; #37 maps 23 requirements and provides review handoff; #38 checks sending prerequisites during collector outage and recovery. VCP guide #3 and #4 distinguish illustrative SDK names from the available SDK, and historical certification procedures from VSO’s current role.

Runtime checks passed first-party Python 3.12 / 3.14 CI. Saved-evidence rechecks are not fresh runs; automated reports are not human review. Proposals and extensions remain unadopted; independent third-party reproduction is not established. Erasure is not universal deletion; collector evidence is not independent observation or all-route bypass prevention. Historical tests and fixed distributions remain intact.

Read the eight updates and limits →

30 October: evidence assessment, uploads and authorization research

VAP PR #30 proposes evidence bindings; #31 provides an offline assessor (20 test methods; 23/23 matching scenarios); #32 exercises loopback uploads (39 methods; 17/17 upload cases and 17/17 evidence rechecks match). Evidect #20 adds a synthetic authorization gate (37 tests; 18 scenarios); #21 adds optional SQLite replay reservations (33 new tests; 70 gate-specific tests and four process scenarios). Counts remain separate by scope.

All results are first-party CI checks. Upload case UL-14 shows a successful bypass outside the gate, not OS isolation or production bypass prevention. Experimental extensions remain unadopted; independent third-party reproduction is not established. Historical 159 / 26 and 306 / 140 test records and fixed distributions are preserved.

Read the five updates and limits →

VSO updates SDK evidence checks, the Explorer demo and Evidect reproduction guidance

29 October: stricter SDK inputs, an API-connected signed-evidence Explorer route and Evidect reproduction guidance are available. The legacy dashboard and historical 159 / 26 and 306 / 140 test records remain separate. Experimental extensions are unadopted; independent reproduction is not established.

The new demo uses signed synthetic evidence; intact full evidence is still INDETERMINATE. The API-disconnected / unsigned description in the 27 October record applies to the legacy dashboard.

Read the update and limits → · SDK Evidence Demo · Evidect

VSO updates VAP-LAP references, VAP errata and SDK boundary checks

28 October: VAP-LAP proposal links, current Draft 3 errata/hash references and SDK boundary checks are available. SDK first-party CI passed 306 Python and 140 TypeScript tests per matrix job, including 13 shared cases and 15 focused tests per language. The Explorer 159-test notice below is retained separately. Experimental extensions remain unadopted; independent reproduction has not been established.

Read the update and limits →

Home / Specification

Current VAP source reference · 2026-11-12

The source remains v1.2.0 Draft 3 and includes editorial erratum E-2026-11-05 and the G-2 withdrawal. Its SHA-256 is unchanged from the main site’s 11 November reference; the VAP subsite previously displayed the older 3 November digest. This source check does not adopt a proposal. Experimental extensions remain unadopted; independent third-party reproduction is not established.

VAP v1.2.0 Draft 3 · e00ee5a · SHA-256 history

SHA-256: a35d5e082fd4b2f9921a78b1a71df263018320375e09fa18cf81533a18a19998

Historical VAP source reference · 2026-11-03

VAP PR #45 adds the previously published commercial-claim withdrawal directly beside historical G-2 and records the disposition. The original historical text is retained. The withdrawn contract figure is not a verified commercial result and is not replaced with zero. The document remains v1.2.0 Draft 3; no normative requirement changed. The source correction is dated 2 November; this website notice is dated 3 November, after the merge.

VAP v1.2.0 Draft 3 · 5ed1084 · Digest history

SHA-256: 4eef5270fbd08bc1ffbae8f08d8bcb41a7f961a70b4a8e4edb75d6f2af280583

VAP source/hash, public Explorer and 23 merged corrections. First-party checks are distinct from independent reproduction; experimental extensions remain unadopted.

Previous website reference (28 October 2026) — informative errata of 25 and 27 October 2026. Draft 3 remains in effect as a draft; experimental repairs are unadopted and independent reproduction is not established.

VAP v1.2.0 Draft 3 · historical text · Errata / hash history

SHA-256: 429ae7765438625adfb5ba66c380b4c2108de3c1be78e376422f25348346e415

VSO-VAP-SPEC-001 · v1.2.0 Draft 3 · status correction 2026-10-24 · CC BY 4.0

Historical reference (24 October 2026) SHA-256: 24a961ddd419298d02b2cc2b8015d6e058501a2411ffa9a744a218bf110ac6bf

VAP Specification v1.2

VAP is a cross-domain upper-level framework defining the structural requirements for cryptographically verifiable decision provenance in high-risk AI systems. Its scope is explicit: domains where system failure can cause irreversible harm to human life, societal infrastructure, or democratic institutions — finance, healthcare, transportation, energy, and public policy as mandatory application domains, with content/creative AI and capture provenance as profile-served domains.

The full text is maintained on GitHub; this page summarizes the load-bearing constructs. It is a summary, not the normative text.

What v1.2 changed

Version 1.2 is a framework-compatible / certification-stricter revision: all v1.1-conformant event data remains valid (event-data-compatible at the wire level; zero wire-format breaking changes). It (a) abstracts “Hash Chain MUST” into Cryptographic Sequence Verifiability MUST, (b) elevates external anchoring to a MUST at all conformance levels, and (c) promotes shared requirements — anchored-batch completeness verification, policy identification, ERASURE (crypto-shredding) events, bounded RECOVERY operations, cross-party provenance (XREF), and SCITT/COSE interoperability — into the Shared Assurance Core. Rationale and migration plan: VSO-VAP-CHANGE-001.

Core invariants (selection)

IDRequirement
INT-006Merkle batching and external anchoring of signed roots at all conformance levels; lightweight anchor targets (e.g., public RFC 3161 time-stamping) explicitly acceptable at the lowest levels.
INT-007A documented anchor continuity plan: fallback anchor target, failover procedure with a bounded migration window, and retained AnchorRecords sufficient to verify historical anchors.
INT-008Completeness Invariant. A third party must be able to verify, for any anchored batch, that no events within the batch’s declared scope were omitted after anchoring (omission / split-view detection). Each AnchorRecord binds event count, first/last event identifiers, and the governing policy identifier.
INT-009Recovery operations that alter the effective sequence must be bounded, recorded, and authorized — and are never filtered from anchor batches.

Completeness scope (informative): INT-008 requires post-anchor omission / split-view detection at batch granularity within the declared scope. It does not prove that every real-world event was observed. Actual verification depends on the evidence, authenticated batch metadata and implementation; the VCP examples have known inconsistencies under review.

Missing data: the three tiers

Three-tier missing-data taxonomy (VAP family)
TierConditionStatus under VAP
Tier 1The event was never measured — nothing crossed the observation boundaryPermanently unrecoverable by design. VAP is silent; no mechanism reconstructs it, and no deployment may represent otherwise
Tier 2Measured, but the record was lost before anchoringBounded and disclosed — the gap itself is recorded as an anchor-gap event with its bounds
Tier 3Anchored, then omitted from what is presentedThird-party detectable via the Completeness Invariant (INT-008): omission / split-view detection

Tamper-evidence answers “were these records altered?”; the Completeness Invariant answers “are these all the records?”. Both are required for evidentiary credibility — and Tier 1 remains, deliberately, out of reach. Events never measured are unrecoverable by design.

Evidence Pack

A portable, self-verifying bundle for offline third-party verification: events, Merkle proofs, AnchorRecords, external anchor tokens, and key material references — verifiable without access to the system that produced them.

Versioning and change control

Version numbers identify documents; SHA-256 digests fix exact meaning. Approvals attach to digests, not version labels. Pre-release review findings fold directly into drafts with disposition records; approved documents are left untouched, and conflicts are resolved by precedence rules in the specification. Change control: VSO-VAP-CHANGE-001 (normative annex).

§1.6 Legal Scope and Non-Guarantee Statement (Normative)

“VAP and its domain profiles define mechanisms for producing cryptographically verifiable evidence of AI system decisions. Conformance to VAP or any profile: (a) does not constitute compliance with the EU AI Act, GDPR, MiFID II/III, CAT Rule 613, NIS2, FDA SaMD guidance, or any other law or regulation; (b) does not constitute a legal determination that any technical mechanism (including crypto-shredding) satisfies a specific legal obligation; (c) does not warrant the correctness, fairness, or safety of the underlying AI decisions — only the integrity, completeness (at anchor granularity), and attributability of their records. VAP generates evidence; competent authorities and courts evaluate it.”

— VAP v1.2, §1.6 (profiles SHOULD adopt this statement verbatim)

Acknowledged gaps

Stated in the specification and tracked openly: there is not yet a standardized vocabulary for decision-authority allocation; only the policy identifier is anchored, not the policy document text itself (no cryptographic binding of policy text yet); and the observation boundary imposes hard limits that no future revision can soften — see Tier 1 above.

IETF Internet-Drafts