Independent AI verification methodology
The method behind every investigation, in plain terms. It is designed so that each conclusion can be traced to evidence, challenged by a person and reproduced by someone else.
The method, step by step
Observation
Study the target from the outside using public and observable material, without access to internal systems.
Claim detection
Extract statements that can be tested, with their source, and separate them from opinion.
Evidence
Collect evidence for and against each claim and record how and when it was obtained.
Contradictions
Show evidence that contradicts a claim next to that claim, never apart from it.
Contrast
Put findings to the organization and record its answer as evidence.
Authorization
Ask for internal access only when needed. It is explicit, scoped and revocable.
Internal review
With authorization, verify internal policies, logs and evidence provenance.
Agentic layer
With authorization, verify agents, workflows, prompts, tool calls, retrieval, handoffs and decision paths.
Human review
A person approves or escalates every finding. V1LA shows why.
Integrity
Hash and timestamp evidence so tampering or loss can be detected later.
Traceability
Link every conclusion back to the evidence and the steps that produced it.
Reproducibility
Record enough that an evaluation can be run again and compared.
Verifiable outcome
A report whose every statement can be followed to its evidence.
What each status means
The same statuses, with the same meaning and the same colors, in the product, in reports and on this site.
- Verified
The evidence supports the claim and the integrity checks on that evidence pass.
- Partially verified
Part of the claim holds. Part is unsupported or out of scope.
- Contradicted
Evidence contradicts the claim. The contradicting evidence is shown next to it.
- Insufficient evidence
There is not enough evidence to support or reject the claim.
- Unresolved
Open: waiting for evidence, a response or a review.
- Contrast required
The finding must be put to the organization before it can be settled.
What the method does not claim
A V1LA finding is a statement about evidence: what supports a claim, what contradicts it and what is missing. It is not a legal opinion, a certification or a guarantee of compliance.
When evidence is insufficient, the honest result is INSUFFICIENT EVIDENCE or UNRESOLVED. V1LA does not fill gaps with assumptions.