Sample investigation: how V1LA verifies an AI claim
A complete investigation of a fictional company, from first observation to final report. It shows the real flow of the platform.
The request
A user chose what they wanted V1LA to verify and started the investigation. Nothing else had to be provided.
- Target type
- Company: Company Alpha
- Layer in use
- External / Organizational Layer
Observe
V1LA studies the company from the outside. External evidence collected without any access to internal systems. Each item is hashed and timestamped.
- EV-0001User-facing systemChat assistant session captureHash: 9f3a1c7e52b0d84a6e…
- EV-0002Public documentHelp center article snapshotHash: 4be07d1a9c32f6850a…
- EV-0003Observable outcomesCredit decision flow, timing analysisHash: c81d5e0a7f3b94126a…
- EV-0004Public documentAnnual report, risk sectionHash: 7a2e93c0b5d1f8461c…
- EV-0005Policy statementTrust center retention statementHash: e50b8a42d7c196f30b…
- EV-0006Organizational documentProcurement policy excerptHash: 1d94c7b3e8a05f2690…
Detect
It extracts claims and finds what does not add up. Claims extracted from the company’s own material, each with its source and its current status.
- ContradictedCLM-001All automated credit decisions are reviewed by a human before final approval.Public AI policy, §4.2
- VerifiedCLM-002Customer chat assistant discloses that it is an AI system at the start of each session.Product documentation
- Insufficient evidenceCLM-003Model outputs are logged and retained for 24 months.Trust center statement
- ContradictedCLM-004Training data excludes customer-identifiable information.Investor presentation, p. 17
- Partially verifiedCLM-005Fraud-detection model is re-evaluated against a fixed benchmark every quarter.Annual report, risk section
- VerifiedCLM-006Users can request human escalation at any point in an automated flow.Help center
- UnresolvedCLM-007Third-party models are assessed before production use.Procurement policy
- VerifiedCLM-008Automated decisions can be explained to affected customers on request.Customer terms, §9
Contrast
It puts a contradiction to the organization. One claim in detail: what supports it, what contradicts it, what the organization has answered and what stays locked until authorization is granted.
All automated credit decisions are reviewed by a human before final approval.
Source: Public AI policy, §4.2
- EV-0007AI policy, human review clause
- EV-0003Credit decision flow, timing analysis
- EV-0011Approval notice, no review step
- OrganizationOrganization responseAwaiting response
- InternalInternal decision logsRequires authorization
- Supporting1
- Contradicting2
- OrganizationAwaiting response
- InternalRequires authorization
- Supports
- Contradicts
- Organization response
- Needs authorization
Verify
Each claim gets a status and traceable evidence. Integrity is checked on the evidence, and a person reviews each finding before it enters the report.
Internal Agentic Layer — authorized example
Internal evidence about this claim, such as decision logs, stays locked in this sample. If the organization authorized it, V1LA would verify those records and add them to the same evidence map.
Not activated in this sample.
The result
The report lists each claim with its status and links every conclusion to its evidence. This is the outcome of the sample investigation.
The demo uses the same fictional data. It is a product demonstration, not a customer account.