# AI Transparency & Acceptable Use Policy

**Last Updated:** June 2026

PrivacyGuard AI is committed to the responsible, ethical, and safe deployment of artificial intelligence within the enterprise sector. This Acceptable Use Policy (AUP) outlines the boundaries of acceptable use and our transparency commitments, aligning with the principles of the UK Data Protection Act and the EU Artificial Intelligence Act.

## 1. Transparency of AI Systems
The Platform functions as an orchestration layer connecting Client prompts to advanced language models. 
*   **System Identification:** All automated reports, document generations, or programmatic summaries output by the Platform should be recognized as AI-generated. We recommend users maintain the standardized compliance footers on exported documents to ensure transparency with end-users.
*   **Human-in-the-Loop (HITL) Requirement:** The Platform is designed as an assistive tool, not an autonomous decision-maker. The Client must maintain human oversight for all outputs.

## 2. Acceptable Use Boundaries
To ensure the safety and legality of the Platform, the Client agrees **NOT** to use the Platform for any of the following prohibited activities:

*   **Automated High-Risk Decisions:** Using the Platform to perform automated, unilateral executive actions that bypass human confirmation, specifically involving employment evaluations, tenant screening, automatic financial credit risk assignments, or legal profiling.
*   **Deception & Fraud:** Generating deepfakes, impersonating individuals or organizations, or generating content intended to deceive end-users about the involvement of AI.
*   **Harmful Content:** Generating text that promotes discrimination, violence, illegal acts, or harassment.
*   **Bypassing Security:** Attempting to jailbreak, reverse-engineer, or maliciously manipulate the underlying API models or the Platform's local masking pipeline.

## 3. PII Masking & Data Minimization
While the Platform provides an automated local masking pipeline to redact Personally Identifiable Information (PII) before transmission, the Client remains responsible for practicing data minimization. Clients must not intentionally bypass the masking pipeline to submit raw, unredacted sensitive data (such as health records or full financial histories) into the prompt environment.

## 4. Enforcement and Suspension
PrivacyGuard AI reserves the right to monitor account usage metadata (not the content of masked payloads) for suspicious activity. Violation of this Acceptable Use Policy may result in immediate suspension or termination of the Client's API access and platform account, without liability or refund.