Model & Data Governance Policy

1 · Purpose

This policy governs how Nxt7a manages AI models and the data they use across their lifecycle, so that models are well-documented, data is fit and rightfully used, and performance is monitored.

2 · Scope

Applies to AI and analytical models Nxt7a develops, fine-tunes, or operates as part of the Earth Intelligence platform, and to the datasets used to build, evaluate and run them.

3 · Policy Statements

Lifecycle governance. Models are managed through their lifecycle – design, data preparation, development, validation, deployment, monitoring and retirement – with appropriate review at each stage.

Data for AI. Data used to train, test or run models must be:

• Rightfully sourced – with clear provenance and the rights/consents to use it, including for Indigenous data per the Data Sharing & Indigenous Data Sovereignty Policy.
• Fit for purpose – relevant, sufficiently accurate, and assessed for quality and representativeness.
• Classified and protected – handled per the Data Classification & Retention Policy.

Bias and fairness. Data and models are assessed for bias that could lead to unfair or misleading outcomes, with mitigation where found.

Validation and accuracy. Models are validated against defined criteria before deployment. Given the platform's environmental purpose, biodiversity and geospatial outputs are validated for accuracy, and known limitations documented.

Documentation. Each significant model has documentation covering its purpose, data, assumptions, limitations, evaluation results and intended use – sufficient for oversight and reproducibility.

Monitoring and drift. Deployed models are monitored for performance and drift; material degradation triggers review and, if needed, retraining or withdrawal.

Change control. Material changes to models or their data go through review and, where relevant, an updated impact assessment.

Third-party models. Where Nxt7a uses third-party models (e.g. via Vertex AI or external APIs), it assesses their suitability, terms and data handling, and does not assume accuracy without verification.

4 · Roles & Responsibilities

• Model owners / developers maintain documentation, validation and monitoring.
• AI Governance Lead sets standards and reviews high-impact models.
• Data owners ensure data rights and quality.

5 · Compliance, Monitoring & Breach

Undocumented or unvalidated models in production are a breach. Governance is checked through reviews and audits.

6 · Related Documents

Responsible AI Policy · AI Risk & Impact Assessment Policy · AI Use Guidelines · Data Classification & Retention Policy · Data Sharing & Indigenous Data Sovereignty Policy.

7 · Standards Mapping

• B Corp: Customers.
• ISO/IEC 42001:2023: A.6 (AI system lifecycle); A.7 (data for AI systems).
• ISO/IEC 27001:2022: A.5.12 (classification of information used by models).

8 · Version History