Independent AI verification for mining
Mining companies and their suppliers use AI to predict equipment failures, support safety decisions and prepare sustainability reports. V1LA verifies the AI systems and the claims made about them; it does not use AI to audit mines.
Main risks
Safety decisions
Predictive models treated as reliable for safety without visible validation.
Sustainability reporting
AI-assisted environmental or ESG figures without a traceable method.
Automation with unclear oversight
Operational workflows described as supervised while the observable process is automated.
Supplier claims
Technology vendors describing accuracy or safety benefits without evidence a buyer can inspect.
How AI is used here
- Predictive maintenance of fleets and plant
- Safety monitoring and alerting
- Process optimization and planning support
- Environmental and sustainability data analysis
Start from the outside. Go deeper with authorization.
V1LA can begin without any access to internal systems. Internal verification is a deeper step that only happens with authorization.
External / Organizational Layer
From public material, V1LA can test what an operator or supplier says about its AI.
- Sustainability and safety reports that describe AI use
- Supplier documentation and public case descriptions
- Published claims about safety, efficiency or emissions
- Public statements about human supervision of operations
- Third-party technology providers named or implied
- Inconsistencies between reports, filings and communications
Internal Agentic Layer — with authorization
With authorization, V1LA verifies how AI behaves in operation.
- Model outputs against recorded plant or equipment events
- Alert and escalation logs
- Agentic or automated workflows and their decision paths
- Access permissions to operational systems
- Human intervention records
- Provenance of the data behind reported figures
Authorization is explicit, scoped and revocable.
Example claims
Statements V1LA would extract and test.
Example claims
- “Our predictive maintenance model reduces unplanned downtime.”
- “A person reviews every safety alert before action is taken.”
- “Reported emissions figures are produced and checked with traceable methods.”
Example evidence
- Public sustainability and safety reports
- Supplier documentation on how the model was validated
- Published descriptions of the operational workflow
- With authorization: event logs, alert records and data provenance
Possible contradictions
- A report that states human supervision of every alert, against a described workflow in which the response is automatic.
- A claimed efficiency benefit, against supplier documentation that describes the model as decision support only.
Evidence gaps
- No reference to validation data for the predictive model.
- No description of how reported figures are produced.
- No record of overrides by operators.
Possible findings
Illustrative outcomes. They are examples, not results about any real organization.
- Contradicted
The supervision claim conflicts with the described workflow.
- Insufficient evidence
The downtime claim is supported only by promotional text.
- Unresolved
The emissions method is awaiting a response from the organization.
All examples on this page are illustrative and fictional.
The role of human review
A responsible operations or safety lead reviews each finding before it is reported. V1LA does not decide what is acceptable risk; it shows what the evidence supports.
Traceability and reproducibility
Findings link to the report, document or log they came from, with a hash and timestamp, so the same verification can be repeated and compared later.
What you define and what V1LA does
You define
- Which operator, supplier, system or process to verify
- Which claims matter most
- Whether to authorize access to operational data, and how far
V1LA does
- Finds public reports, filings and documentation
- Extracts testable claims and looks for evidence
- Contrasts findings with the organization
- Reports status and evidence for each claim
Scope
V1LA verifies claims and evidence. It does not replace safety certification, engineering review or regulatory compliance work.
What do you want V1LA to verify?
Describe the company, system, process or AI agent. V1LA does the discovery.
Start investigation