AI audit trails for verifiable records of decisions and actions.
VeritasChain Standards Organization (VSO) develops VAP, a framework for shared requirements across domains, and domain-specific specifications including VeritasChain Protocol (VCP) for finance.
From AI activity to verifiable recordsVAP defines shared requirements. Domain profiles turn them into specifications for each field.
Capture decisions, actions and their context.
Use cryptographic proofs to check record integrity.
Let third parties examine the recorded evidence.
VAP — Verifiable AI Provenance Framework defines common requirements for AI decision records, including integrity, provenance and traceability. Implementations use domain-specific profiles.
VeritasChain Protocol (VCP) applies the framework to algorithmic and AI-driven trading. Explore the specification, reference implementations and verification demo.
Specifications developed and maintained by VeritasChain Standards Organization (VSO)
Explore domain profiles and their status → · About VSO →
Practical guidance for teams evaluating AI agents, record integrity and independent verification.
Connect responsibilities, risk reviews and incident response to records that can be examined.
AI governance: from policy to evidence →Design an AI agent audit log and understand signatures, anchors and verification limits.
AI audit trails: what to record and verify →Trace one operation from delegated permission to submitted evidence, with an editable evaluation checklist.
AI agent audit logs: authority, tool calls and unknown outcomes →Choose a practical guide · Browse all VSO specifications and resources →
What was in scope, what was observed, and why was it judged a problem? An application proposal that combines OAP and DAP so the research process can be checked from evidence.
OAP and DAP are Working Drafts registered in VAP v1.2. The proposal adds no profile and sets no requirements, and it states what the evidence alone cannot guarantee.
Read the design proposal →FOUNDER / VSO
Founder, VeritasChain Standards Organization (VSO)
Tokachi Kamimura founded the VeritasChain Standards Organization (VSO). Through VAP and VeritasChain Protocol (VCP), he works on open standards for verifiable records of AI and algorithmic systems.
Explore his research, technical articles, and development updates on LinkedIn.
View LinkedIn profileCorrection — 24 October 2026: the previously displayed totals of 67 authorities and 50 jurisdictions are withdrawn from current promotional claims pending reconciliation with submission records. This is not a finding of zero submissions.
The map and list include historical outreach records under review. The 26 October 2026 announcements describe 26 individually recorded country contacts: 25 sent emails and one acknowledged form submission. Sending, receipt, technical engagement and approval are distinct evidence states.
Submission, receipt, technical response and approval are distinct evidence states.
Country and authority names appear below. Colors indicate outreach records, not receipt or approval. Natural Earth
Outreach records are being reconciled before totals are restated
Spanning Americas, Europe, Asia-Pacific, Middle East, and Africa
Evidect #30–#34 add synthetic permission and receiver-state experiments. Latest first-party CI passed 306 tests per Ubuntu/macOS job. Extensions remain unadopted; independent…
Evidect #28 adds receiver verification; #29 documents incident sources and scope limits. First-party CI passed 137 tests and eight synthetic cases per OS. Extensions remain…
Three research areas: Evidect post-approval checks and process observations, VAP/Evidect retention review, and Python/Node interoperability. First-party CI only; extensions remain…
VCP / VSO · 26 October 2026
First-party PoC · legacy v1.1-labelled evidence · no tier, VAP conformance or certification established
Report valid, invalid and indeterminate checks separately; missing inputs do not hide independent mismatches.
20 SIG + 20 EXE. Legacy comparisons match the selected reference; overall evidence remains INDETERMINATE.
The published cc695fe procedure completed from fresh clone to final assessment in both first-party Python environments. Its fixed 809 + 15 cases are repeated checks; independent reproduction remains unestablished.
Implementation extension — unadopted; independent reproduction not established
Read the update and reproduction instructions →
A local-only, audit-safe reference implementation
for running the AI Decision Auditability Benchmark
and exporting an Audit Evidence Pack.
VAP Scorecard Explorer is a reference implementation of the AI Decision Auditability Benchmark (10 criteria / 20 points).
It allows audit and assurance teams to:
All processing runs locally.
No network communication. No external APIs. No analytics.
Benchmark specification and scoring criteria are published openly as the canonical reference.
A demonstration dashboard, running on simulated sample data, that shows how independent, cryptographic verification of VCP records works.
No live trading system is connected. The Explorer API source is public. Ordinary API routes return mock data; the separate SDK Evidence Demo is available in the public Explorer. The production URLs and SDK-spec workflow illustrated here are not an available production service. Authentication middleware is not implemented in the published API.
One-click verification of any trade lifecycle
Merkle proofs, hash-chain validation, and timestamp verification are performed entirely client-side — no server trust required.
Essential technical documentation and standards for developers, regulators, and integrators
RESTful API documentation for VCP event queries and verification
View API DocsComplete protocol specification with data models and integrity layers
Read SpecificationTechnical conformance certification scheme for VCP implementations; no independent CAB is currently registered
Learn About CertificationOpen-source repositories including specifications, the SDK specification and reference implementations
Explore RepositoriesStart building with VCP — Explore the API documentation, the SDK specification, and reference implementations
VeritasChain Protocol (VCP) is an open standard specification for recording the "decision-making" and "execution results" of algorithmic and AI-driven trading — including AI model versions, decision factors, time-synchronized event lifecycles, and cryptographically hashed audit trails — in a tamper-evident, append-only and verifiable format. VCP provides a cryptographically secured chain of evidence, designed to support compliance efforts with international regulations including MiFID II, EU AI Act, and CAT (Consolidated Audit Trail), making it a valuable tool for RegTech, Audit, and Compliance. VCP verifies the integrity and attribution of recorded events — it does not certify the correctness or safety of the underlying AI or trading decisions.
In addition to the order and execution data that FIX messages carry, VCP defines records for:
FIX standardises the messages exchanged between counterparties; it does not define a record of a system's internal decision factors or risk state.
This makes VCP relevant to evidence requirements under MiFID II RTS 25, EU AI Act Article 12, and next-generation RegTech.
VSO's position is that oversight of algorithmic and AI-driven trading should rest on records that can be verified independently. The rules below require record-keeping, event logging or clock synchronisation; none of them mandates cryptographic audit trails or refers to VCP.
Requires business clocks traceable to UTC, with specified accuracy, for trading venues and their members and participants in the EU (Delegated Regulation (EU) 2025/1155, Arts. 11–16; replaced RTS 25 on 2 March 2026)
Requires automatic event logging (Art. 12) and human oversight (Art. 14) for high-risk AI systems; Annex III obligations apply from 2 December 2027
U.S. SEC's Consolidated Audit Trail requires granular trade lifecycle tracking
VCP provides a single open standard that produces evidence relevant to all these requirements, reducing fragmentation across jurisdictions — by design.
Tamper-evident audit trails using Merkle trees (RFC 6962 domain-separated hashing) and external anchoring; hash chains are optional
Ed25519 by default, upgradeable to post-quantum algorithms including ML-DSA (FIPS 204)
Each signature carries its algorithm identifier, so algorithms can be changed. Post-quantum signatures are EXPERIMENTAL in VCP v1.2.0 GA (tag v1.2.0, commit b23acb266f88df81d960d36d58fb64ceaa0dd5ed; GA merged 2026-10-08 JST; fix merged 2026-10-09 JST) and are not a certification requirement. This is the VCP release status, separate from VAP v1.2.0 Draft 3.
Designed to produce evidence relevant to MiFID II RTS 25, EU AI Act, GDPR, CAT Rule 613, and international reporting
Compliance tiers for HFT (Platinum), Institutional (Gold), and Retail (Silver).
Note: Silver Tier relies on system time (Best-effort) and delegated signatures. It provides transparency, not regulatory-grade evidence.
Side-car integration with existing FIX engines — no change to trading gateways required
Fully open specification with public reference implementations. Designed for long-term interoperability across brokers, exchanges, and regulators.
Split-View Attack and Omission Attack resistance (v1.1). Designed so that removing events from an anchored batch, or showing different records to different parties, is detectable.
Records algorithm identity and version, a model hash, decision factors and approval records (VCP-GOV). Where a system is classified as high-risk under the EU AI Act, such records are evidence relevant to Article 12 record-keeping; algorithmic trading is not listed in Annex III.
Audit-trail verification, algorithmic decision reconstruction, and cross-border regulatory coordination
Best-execution evidence, trade lifecycle verification, and verifiable records as input to regulatory reporting
Transparent order matching, CEX audit trails, and cross-venue surveillance
Tamper-evident audit trails for high-frequency and algorithmic trading systems
Verifiable trading records for payout and performance disputes
Merkle root anchoring to public blockchains
Integrate VCP into your trading infrastructure using the open specification, the SDK specification and the reference implementations
View Specification (Available Now) →Deploy VCP as a sidecar to existing FIX engines to support regulatory compliance efforts
Contact Us →Industry-specific profiles for cryptographically verifiable AI audit evidence
VAP (Verifiable AI Provenance Framework) — which VSO describes as "the flight recorder for AI" — is the cross-domain upper-level metaframework (current: v1.2.0 Draft 3) that defines the structural requirements for cryptographically verifiable AI decision provenance. Its scope is deliberately strict: domains where system failure could cause serious, irreversible harm to human life, societal infrastructure, or democratic institutions. Finance, healthcare, transportation, energy, and public policy are VAP's mandatory application domains; content/creative AI and capture provenance are served by released profiles. VAP itself is not an implementation — domain profiles apply its requirements to a sector. Status in the VAP registry (§4.5.2): VCP v1.2 Released; CAP v1.0 Released; CPP v1.4 Released; OAP v0.1.1 Working Draft; MAP v0.1.2 Working Draft; PAP Partial Draft (Track B v0.2.0); IAP Announced; DVP, AAP and EIP Planned (no specification document exists for IAP, DVP, AAP or EIP). Read the VAP Framework overview →
Explore VAP-LAP (Legal AI Profile), proposed in an individual Internet-Draft, with AILEX as a product and use-case reference.
VeritasChain Co., Ltd. is a technology company developing open cryptographic audit and provenance standards for algorithmic and AI-driven trading, enabling regulators, institutions, and markets to verify—not merely trust—system behavior.
VAP (Verifiable AI Provenance Framework) defines shared requirements for verifiable records of AI decisions and actions. The published framework is v1.2.0 Draft 3; it is a draft, not a certification of an AI system.
VAP sets cross-domain requirements. VeritasChain Protocol (VCP) is the financial domain profile for algorithmic and AI-driven trading records. Other profiles address different domains.
Cryptographic evidence helps verify the integrity and provenance of recorded events. It does not establish that an AI decision was correct, safe or legally compliant.
Read the VAP framework for the shared requirements. For trading systems, explore the VCP specification and the verification demo using sample data.
Explore VAP Read the VCP specification Try the verification demo
For media inquiries, partnership opportunities, technical collaboration, or standardization initiative participation, please contact us:
VSO does not provide financial services or regulatory approvals.
