MAGNAVIS AI and Automation Policy

Version 1.2 — Effective September 15, 2026

1. Scope

MAGNAVIS uses deterministic rules, scores, search, and local AI-assisted functions over synthetic records in the current Beta. These functions have different limitations and are not autonomous hiring agents. External AI processing depends on the configured provider, access controls and request. Administrator configuration shows which providers are enabled; prompt history records the provider used for each request. Do not assume a provider is disabled from this policy alone. Use requires separate authorization, configuration, and verification before use.

2. Human responsibility

Outputs support, not replace, qualified human judgment. Users must review source data and material outputs, investigate errors, document independent decision reasons, and remain able to disregard, correct, or stop an output. No AI-assisted output may be the sole basis for a decision producing legal or similarly significant effects.

3. Fairness

MAGNAVIS policy prohibits protected characteristics as scoring or decision criteria unless counsel confirms a lawful obligation or exception. This policy prohibition is not a representation that every possible source field has been independently validated, and deterministic design does not eliminate biased, incomplete, or incorrect source data or outcomes. MAGNAVIS does not claim recurring bias audits, independent validation, or nondiscrimination certification unless a current, scoped report is supplied for the applicable feature.

4. Explanations and accuracy

Where a feature supports it, MAGNAVIS displays factors or source information intended to help an authorized user review an output. This does not disclose source code, security-sensitive or third-party confidential information, or guarantee that an external model can fully explain its output. Generative text may hallucinate, omit context, or reproduce source errors.

5. External provider controls

External processing is permitted only for the provider approved for the applicable environment and only after commercial terms, DPA, retention, transfer, and subprocessor requirements are recorded. MAGNAVIS does not opt Customer Data into general model training without separate express authorization. Silent failover to an unapproved external provider is prohibited; local or rule-based degradation is preferred.

6. Customer duties

Customers must configure authorized users and purposes, minimize submitted data, give required notices, honor individual rights, train operators, monitor results, and comply with employment, privacy, AI, communications, and anti-discrimination law. The current synthetic Beta is not authorized for EEA/UK regulated recruitment use; any such launch requires the applicable assessment, contract, and verified controls.

7. Feedback and incidents

Report suspected harmful, discriminatory, inaccurate, or privacy-affecting output to info@middletonadvisorygroup.com. Preserve relevant input, output, model/provider, time, user, and decision context without unnecessarily duplicating personal data.