Interpolgar Responsible AI and transparency
Last updated: October 13, 2026
1. Our approach
Interpolgar Nexus provides explainable decision support. Today its scores and alerts are based on transparent rules and formulas that each insurer can configure; they are not black-box models or machine learning trained on customer data.
2. Explainability
Every recommendation shows the factors, points and rules that produced it, the rule version in force and the evidence. In credit, the system generates the principal reasons for a denial to support the adverse-action notices the law requires. The audit log preserves traceability and cannot be altered without leaving a trace.
3. Human oversight
The system recommends and a person decides. Cases with critical findings, high impact or suspected fraud require review and a written justification. We recommend that customers keep a qualified person in the decision loop.
4. Data and training
We do not use customer data to train our own or third-party AI models. If we ever offer machine-learning features, it will be with express consent by contract, documented governance and prior testing.
5. Fairness and non-discrimination
Rules are designed without using protected characteristics as criteria. Customers are responsible for assessing the disparate impact of their own rules and thresholds; the platform lets them simulate changes before publishing. We are developing fairness reports to make those tests easier.
6. Limits
Recommendations may contain errors or false positives and do not replace professional judgment or legal, medical, financial or actuarial advice. We are not a consumer reporting agency. Poor-quality input data produces poor-quality results.
7. Continuous improvement and reports
We review rules and their performance with customers during pilots and record changes with versioning and rollback. If you believe a decision assisted by the Service is wrong, contact the organization that made it; it can ask us for the decision trace. Contact: info@interpolgarnexus.com.