AI Governance Policy
Effective Date: September 21, 2026 | Aligned with NIST AI RMF 1.0 and ISO/IEC 42001 principles
1. Purpose and Scope
This AI Governance Policy describes how Project Research LLC designs, tests, and maintains the AI/machine-learning forecasting components of ClusterCollie AI. It applies to the forecasting engine's data handling, model behavior, human oversight expectations, and disclosure practices. This policy is informed by the structure of the NIST AI Risk Management Framework (AI RMF 1.0), which organizes AI governance around four functions: Govern, Map, Measure, and Manage.
2. Govern: Accountability and Oversight
- Project Research LLC maintains ownership and accountability for the design, testing, and disclosure of ClusterCollie AI's forecasting models.
- All AI-generated forecasts are clearly labeled as such within the application and on this website via a persistent AI Disclosure notice.
- Beta status is disclosed prominently so users understand the current maturity level of the forecasting models.
3. Map: Context and Risk Identification
ClusterCollie AI's forecasting is scoped narrowly to Nutanix infrastructure capacity trend projection — it is not used for decisions affecting employment, credit, health, or other high-risk domains under frameworks such as the EU AI Act. Identified risks include:
| Risk | Mitigation |
|---|---|
| Forecast inaccuracy from limited historical data | Confidence intervals widen automatically for clusters with short data history; users are prompted to validate against live Prism Central data. |
| Over-reliance on automated output | Terms of Use and in-app messaging require human review before procurement/budget decisions. |
| Data exposure of cluster credentials | Credentials and telemetry are processed locally by default and are not transmitted to Project Research LLC servers. |
| Model drift as usage patterns change | Forecasting parameters are reviewed and updated as part of ongoing beta testing. |
4. Measure: Evaluation and Monitoring
- Forecast outputs are evaluated for validity and reliability against held-out historical data during development.
- Confidence bands are calculated based on historical volatility, providing a transparency signal alongside every projection.
- User-reported forecast anomalies submitted via the Contact page are logged and reviewed to identify systematic model issues.
5. Manage: Risk Response and Continuous Improvement
- Known limitations are documented publicly in the White Paper rather than obscured.
- Beta releases are versioned, and material changes to forecasting logic are noted in release documentation.
- Users can report issues, including suspected bias or systematic forecast error, via the Contact page; reports inform prioritization of fixes.
6. Trustworthy AI Characteristics
Consistent with NIST's trustworthy AI characteristics, ClusterCollie AI is designed to be:
Valid & Reliable
Forecasts are back-tested against historical data before release.
Transparent
AI-generated content and methodology are disclosed on every page and in the white paper.
Explainable
Forecasts show trend direction, confidence bands, and time-to-threshold rather than opaque scores.
Privacy-Enhanced
Cluster telemetry processing defaults to local execution; no forced cloud transmission.
Secure & Resilient
API credentials are handled via local environment configuration, not hardcoded or logged.
Accountable
Project Research LLC retains named responsibility for model behavior and disclosure accuracy.
7. Human Oversight Requirement
ClusterCollie AI forecasts are decision-support tools, not autonomous decision-makers. A qualified human (infrastructure architect, capacity planner, or equivalent) must review forecast output before it informs procurement, budgeting, or workload migration decisions. This requirement is reinforced in the Terms of Use and in the AI Disclosure shown on every page.
8. Policy Review
This AI Governance Policy is reviewed periodically as the forecasting engine evolves and as external frameworks (including NIST AI RMF updates and applicable state or federal AI regulation) develop. Questions about this policy can be directed to admin@projectresearchllc.com.