AI systems intelligence & governance
Know your AI systems. Prove what you know.
AI systems are scattered across applications, infrastructure, repositories and documentation. Verotan brings their evidence together, identifies what can be established, and makes the gaps impossible to ignore. The goal: less manual inventory, more defensible governance.
Working preview Evidence-backed assessments. Automated discovery and continuous monitoring are in development.
Verotan is an AI systems intelligence and governance platform for organisations that build, operate or depend on AI.
An AI inventory shouldn't depend on someone remembering every system, integration and data flow. Verotan is designed to collect evidence from authorised sources, reconstruct the relationships between systems, and identify what still requires human confirmation.
Every assertion retains its basis. Every assessment identifies the evidence it used. And when something changes, the relevant conclusions can be revisited.
Five outcomes. No invented certainty.
Every requirement receives one of five explicit outcomes. Missing evidence never becomes assumed compliance, and uncertainty is never silently converted into a pass.
-
Met
Available evidence satisfies the requirement.
-
Partially met
Some obligations are supported; others remain unmet or insufficiently established.
-
Not met
Evidence establishes that the requirement is not satisfied.
-
Not applicable
A documented applicability rule excludes this requirement.
-
Unable to determine
Evidence is insufficient, stale or conflicting.
Every finding can be traced
A status on its own is not governance. Each finding carries the fact it consulted, the basis for that fact, the source behind it, and the version of the control set the judgement was made against.
- Basis is recorded Observed from a source you authorised, attested by a person, or inferred. AI-generated conclusions remain identifiable as inferences until independently supported or explicitly confirmed.
- Attestation is attributed, not verified A statement is recorded as an accountable person's claim. It is never presented as an independent operational observation of the system's behaviour.
- Counter-evidence is kept If two sources disagree, both are retained and the conflict becomes a review task rather than a silent overwrite.
- Review does not rewrite A reviewer can record a decision against a finding. The engine's original output stays visible underneath the decision.
Human review before decisions take effect
Customer attestation indicates that outputs can take effect without human review. No independent operational verification is recorded.
Training data provenance
The available evidence does not establish whether personal information was used in training or fine-tuning. Additional information is required before this control can be evaluated.
Draft Verotan baseline controls, version 2026.11-draft, evaluated by control-evaluator/3. This is a draft control set, not a government certification or an authoritative determination of regulatory compliance.
When something changes, know what it affects. In development
A model changes. An application gains a capability. A data source moves. A governance requirement is revised.
Verotan is being designed to connect those changes to the systems, controls and findings they affect, so teams can reassess what matters instead of starting every review from scratch.
- Requirement revised Identify which assessed systems the change reaches before anyone re-opens a review.
- System changed Identify which controls, findings and evidence have gone stale as a result.
- Reassess what matters Target the affected record instead of re-running the whole programme.
How Verotan is designed to work Product direction
The long-term workflow is discovery-led: the product does the investigative work rather than providing nicer forms to fill in. Most of this is not built yet, and the tags below say what is.
-
Connect In development
Authorise repositories, cloud infrastructure, applications, documentation and telemetry.
-
Discover In development
Identify AI systems, dependencies, data relationships and supporting evidence.
-
Complete Planned
Present inferred facts for confirmation and ask targeted questions about what cannot be established automatically.
-
Assess Available now
Evaluate evidence against versioned governance controls, with explainable findings and explicit uncertainty.
-
Monitor Planned
Reassess affected systems when either the systems or their governing requirements change.
We ingest what you authorise, infer what we can substantiate, and ask you for the rest.
What you can try today Available now
Start with a working assessment. Register a sample AI system, provide six answers, and see how Verotan produces traceable findings from a versioned control set. This is one part of the larger discovery and governance platform under development.
-
Register a system
Name it, describe what it decides and who it affects, and note where it runs. This becomes the snapshot every finding is pinned to.
-
Answer six questions
Each answer is stored as an attestation, attributed to the person who gave it and dated. Leave a question blank and it stays blank.
-
Evaluate
The control set runs as typed rules, not as a prompt. The same inputs always give the same findings, and each explains what it consulted.
-
Review the findings
Record a decision and a justification against any finding. The engine's original status remains visible next to it.
Evidence you can inspect. Decisions you can defend.
Verotan is designed around explicit access, traceable evidence and accountable human review. Observations, inferences and attestations retain their sources. Conflicting information is exposed rather than erased. Decisions remain reviewable.
Try the assessment engine
Use a fictional or non-sensitive system to explore evidence-backed findings, explicit uncertainty and the draft baseline controls.
The baseline control set is a draft authored by Verotan. It is a starting point for review, not a certification.