AI Transparency Policy

Edexia ensures clarity, user control, and responsible deployment of Artificial Intelligence (AI) in its educational tools. This document outlines how our AI functions, what data it uses, its limitations, and how users are informed and given control over updates and model changes.

1. How Our AI Works

Edexia’s AI assists educators by interpreting grading criteria and analyzing student submissions. It uses OpenAI’s GPT-5 to generate draft feedback and explanations.

  • Data sent to the model is de-identified for privacy.
  • The AI may optionally use web search to access up-to-date information.
  • Explanations are provided so users can see why feedback was generated.

2. Limitations

AI responses are not guaranteed to be correct or complete. Educators must review all output before using it.

  • Results may vary due to incomplete context or input ambiguity.
  • GPT-5’s knowledge base may contain biases or inaccuracies.
  • Edexia’s AI is a support tool, not an autonomous grader.
  • On-screen disclaimers remind users to verify accuracy.

3. Versioning and Updates

AI systems and features are versioned, labeled, and communicated clearly.

  • New features are tagged as New or Beta.
  • Current version: GPT-5.
  • Silent model switches do not occur during use.
  • Notifications are issued for significant upgrades.

4. User Control

Users may opt in or out of new AI models and functions.

  • Opt-In: Experimental tools remain optional until stable.
  • Opt-Out: To disable new models or revert versions, email admin@edexia.ai.
  • Institutions may coordinate phased adoption directly with Edexia.

5. Change Logs and Notifications

All major updates are documented and communicated.

  • Change logs list AI and feature modifications.
  • Users may request detailed history via admin@edexia.ai.
  • In-app and email notifications announce major changes.

6. Contact

Questions, feedback, opt-out requests, or changelog inquiries:
admin@edexia.ai

Edexia deploys AI with transparency, explicit versioning, and user autonomy to ensure reliability and trust in educational settings.