VSO publishes seven VAP payment-evidence research and AI design updates
VeritasChain Standards Organization (VSO), operated by VeritasChain Co., Ltd., summarizes seven VAP research and design updates merged on 6 November. Executed synthetic experiments are distinguished from unexecuted proposals.
Unadopted research and design proposals
The experimental extensions and design proposals remain unadopted. Independent third-party reproduction has not been established. First-party CI success and merging research do not establish adoption, full normative conformance, certification or production suitability. No real bank, real funds or production credentials were used.
Payment evidence: record integrity is separate from payment authority (#52, #54)
Proposal #52 separates cryptographic integrity, business authorization, execution matching and observed control state. The offline experiment in #54 provides 24 synthetic fixtures mapped to PAY-01–08 and 57 unit tests. It verifies real Ed25519 signatures and bindings to original bytes, retaining unknown judgments when evidence is insufficient. This does not execute the complete proposed acceptance contract, connect to a bank or demonstrate an admission gate.
Dispatch and recovery: no automatic payment resend after a lost response (#55)
A loopback mock provider and SQLite journal record consumption of the first dispatch opportunity before permitting one POST. After response loss or restart, recovery uses read-only GET reconciliation and retains an unknown outcome where necessary. There are 28 unit tests and a demo that reconciles one POST with one mock balance effect. Business authorization is assumed by this experiment; it does not establish generic exactly-once execution or OS isolation.
Two-person approval: two keys for one person are not two approvers (#56)
Configured trusted keys and person mappings, the exact request, policy and supplied signed-status freshness are checked when the dispatch claim is recorded. Twenty tests and a nine-stage demo cover rejection of same-person approvals, changed amounts, revocation, missing evidence and stale status. This is a bounded experiment with supplied signed snapshots, not real-person authentication or a guarantee of the latest revocation state at the instant of effect.
Collection: distinguish pre-dispatch custody from post-effect evidence status (#57)
A separate mock collector process under the same operator stores original admission and receipt bytes. Initial dispatch requires a signed ACK, nonce-bound readback and locally recorded prerequisites. Twenty-five unit tests and five demo scenarios with 13 stages exercise collector stoppage, ACK loss, lost payment responses and collection failure after an effect. Recovery only collects or reconciles original evidence; it cannot resend payment. SETTLED or FAILED outcomes remain separate from subsequently unknown custody. A separate process is not independent custody or third-party timestamping.
School AI and consumer agents: design documents only (#53, #58)
School-AI proposal #53 separates requests and responses, risk judgments, alert attempts, provider acceptance, delivery, human acknowledgment, configuration and observed effects. No school integration or new runtime experiment was executed. Consumer-agent proposal #58 organizes principals, declared commercial relationships, captured inputs and candidates, disclosures, authorized actions, outcomes, custody and deletion evidence. CA-01–08 are eight unexecuted review cases, not passed tests. These documents do not establish deployed product features or new legal obligations.
First-party CI on a fixed revision
Four workflows succeeded on the post-merge payment-research revision ce9c968951c0c92a3b5cf9d5a9a587506f8a700d (#57): 57 offline-evidence tests, 28 dispatch/recovery tests, 20 authorization tests and 25 collector tests. Repeated normal/optimized runs and overlapping workflow reruns are not additional unique tests. The subsequent #58 revision d97980f changes documents only and has no new runtime CI result at that HEAD. Preparing this announcement did not rerun the research suites.
All experiments use first-party synthetic environments. A holder of raw mock-provider credentials can bypass the wrapper. Complete route coverage, independent custody, external anchoring, OS privilege separation and full proposal conformance are not established. A signed ACK does not prove truth, completeness or continuing retention.
Merged primary sources
Preserving revision-specific records
The 5 November site-main CI results, the Explorer record at 159 tests and the fixed reproduction kit remain linked to their original revisions and dates. This website update only adds announcements; it does not alter current normative requirements, existing implementations, signed data, fixed distributions or K3.