AI Acceptable Use Policy

1. The Core Principle: "Human in the Loop"

The Same Page uses advanced artificial intelligence to help you explore workplace rights and structure accommodation requests. However, Your Digital Advocate is a "Strategic Co-pilot," not an autonomous decision-maker.

• You Are the Author: You agree that you are the final author of any content generated by our platform.
• Mandatory Review: You must review every draft for accuracy, tone, context, and potential bias before sharing it with your employer. You are responsible for ensuring the final request is fair and accurately represents your lived experience.
• Non-Directive AI: The AI will explicitly define its role to help structure thoughts, but it is strictly forbidden from telling you what to do.
• The Privacy Firewall: Your exploration within the platform is completely isolated. Managers and HR Admins have absolutely zero visibility into your AI interactions or drafts until you explicitly choose to generate and share an authorized Accommodation Request.

2. Privacy & Data Architecture Guarantees

Our AI Governance Framework is built on Privacy-by-Design principles, ensuring your data remains protected at the architectural level. When you use the AI, you are protected by the following structural guarantees:

• Strict Data Residency: 100% of our infrastructure, database, and AI processing is strictly locked to the AWS ca-central-1 (Montreal) region to ensure PIPEDA compliance.
• Zero-Day Retention: AWS Bedrock "Model Invocation Logging" is strictly disabled. AWS does not store your prompt data, nor is any user data used to train foundational AI models.
• The Pseudonymization Layer: The AI never knows your real name or email. Every standard user is assigned a unique, 8-character alphanumeric pseudonym.
• Temporary Document Memory: Documents you upload to the Discovery Space are read temporarily so the AI can help you, but they are never permanently saved to our databases.
• Dynamic Pseudonym Rotation: To maximize long-term anonymity and prevent tracking, your visible 8-character alphanumeric identifier is dynamically rotated every 90 days. Past data remains securely unlinked from your current identity to downstream corporate actors.
• Mobile & Device Boundaries: Because AdaptAbility is accessible via mobile applications (iOS and Android), you are advised to disable third-party cloud clipboards or predictive text syncing when handling highly sensitive medical context. AdaptAbility explicitly never writes app state data to unencrypted local storage.

3. The "Medical & Legal" Boundary

Your Digital Advocate is grounded in verified legislation, open-access clinical research, and community-based insights rather than the open internet. However, it is not a doctor or a lawyer.

• Verification: Every piece of advice or legal context provided by the Digital Advocate is tied directly to a specific source document using strict APA style citations.
• No Diagnosis: You must not use this tool to diagnose a medical condition for yourself or a colleague.
• No Legal Counsel: Do not use this tool to generate formal legal threats or binding court documents without consulting qualified legal counsel.
• Two-Phase Language Protocol: You may freely interact with the Digital Advocate in your preferred native language during the exploration and brainstorming phase. However, you agree that any final, binding accommodation request outputs must be standardized into English or French to ensure your organization can legally act on them.

4. Scope Management & Prohibited Uses

To ensure the platform remains a dedicated, secure professional tool, the AI employs intelligent scope management. If the Advocate detects queries veering into topics entirely unrelated to the workplace, employment rights, or accommodations, it will gently intercept the conversation and ask you to confirm your use aligns with this AUP.

To keep this platform safe for everyone, you strictly agree NOT to use our AI for any of the following:

• Harmful Content: Generating content that promotes self-harm, suicide, violence, or is discriminatory, racist, sexist, ableist, or harassing.
• Deception: Creating content for the purpose of deceiving others, including deepfakes or impersonating a medical/legal professional.
• Hacking & Manipulation: Attempting to override, trick, or "jailbreak" the AI's safety instructions, or trying to reverse-engineer our proprietary data.
• Privacy Violations: Inputting the Personally Identifiable Information (PII) of other people without explicit consent, or uploading sensitive company secrets (e.g., financial keys, passwords).
• Automated Scraping: Using bots or scripts to harvest data or generate bulk content from the platform.
• Intellectual Property: Submitting copyrighted materials, proprietary code, trade secrets, or confidential third-party intellectual property without explicit authorization.

5. Safety, Safeguarding & The Circuit Breaker

Because AdaptAbility operates as an independent "Privacy Firewall," safeguarding is a critical operational component.

• Distress Scoring: The system actively monitors conversations using a 0-10 internal distress scale.
• Pastoral Support (Level 8): If an interaction reaches a level 8, the AI will deliver a targeted, pastoral message of support to help validate and ground you emotionally.
• Crisis Intercept (Level 9-10): If language indicates a potential life-safety event, the AI instantly intercepts the conversation, provides localized, human-verified resources (such as 988 in Canada), and escalates the alert to a trained human administrator.
• The "Break Glass" Protocol: In a confirmed crisis, a TSP Admin can execute a secure protocol to reveal your identity for emergency escalation. This action generates a permanent, read-only Privacy Audit Log entry requiring an admin justification.

6. Input/Output Ownership & Error Reporting

• Your Rights: You retain ownership of the text you type into the chat. To the fullest extent permitted by law, you own the rights to the text the AI generates for you after you have reviewed and accepted it.
• Handling Hallucinations: AI can sometimes "hallucinate" or generate factually incorrect information. If you spot dangerous or wrong advice, do not use it.
• Feedback Mechanism: Report issues using the feedback tools. To protect your privacy, bug reports and feedback use a secure, untraceable identifier, making it mathematically impossible for admins to reverse-engineer your identity while still allowing us to fix technical errors.
• Security Incident Reporting: If you suspect a data breach, privacy violation, or severe AI malfunction, you must immediately halt use and report the incident via the secure feedback portal to initiate our formal incident response protocol.

7. Enforcement

We reserve the right to monitor usage patterns for safety and policy violations. We employ a stepped approach to enforcement:

• Minor Violations: Unintentional misuse may result in a warning, a temporary pause on your account, and a requirement to review our AI training materials.
• Severe Violations: Intentional abuse, malicious jailbreaking, or generating harmful content will result in immediate account suspension and notification to your organization's HR Administrator.
• Imminent Harm: If we detect content involving imminent harm to life or safety, we will execute our "Break Glass" protocol and report it to relevant authorities or emergency contacts as required by law.
• Appeals: If you believe your account was suspended in error, or if you feel the AI has generated unfairly biased outputs, you may submit a review request to The Same Page Administrative team via our secure feedback portal.

To ensure ongoing compliance with evolving AI governance standards, you may be required to periodically re-acknowledge this AUP or complete updated AI safety training modules to maintain active access to Your Digital Advocate.