AI Transparency

Plain-language explanations about how our AI works, what it sees, and what it never does.

How the AI works

RareJourney's AI companion uses a technique called Retrieval-Augmented Generation (RAG). Here's what that means in plain language:

1

You ask a question

For example: "What foods should I avoid with my medication?"

2

We search your relevant data

The system searches your health profile, uploaded documents, and timeline for relevant information. It also searches our curated library of published research and community insights.

3

We send context to Azure OpenAI

Only the most relevant snippets (not your full records) are sent to Azure OpenAI with your question. This is called "context."

4

The AI generates a response

Azure OpenAI uses the context to generate a personalized, evidence-based answer. The AI doesn't "remember" your past conversations unless they're part of the current chat history.

5

You get a helpful answer

The response is based on your specific health context plus published evidence, making it more relevant than a generic web search.

Why RAG? This approach means the AI isn't making up answers from thin air. It's grounded in your actual data and published research, making responses more accurate and trustworthy.

What data the AI sees

The AI companion only sees relevant snippets of your data when responding to specific questions. It does not have access to your full health record.

Sources the AI can access:

  • •Personal health profile: Diagnosis, medications, symptoms you've logged
  • •Uploaded documents: Relevant excerpts from medical records, test results, care plans (PII redacted)
  • •Timeline entries: Your daily logs and observations relevant to the question
  • •Published research: Curated library of medical literature and rare disease guidance
  • •Community insights: Aggregated, anonymized patterns from other users (opt-in only)
  • •Chat history: Recent conversation context to maintain continuity

Data minimization in action:

If you ask "What foods should I avoid?", the AI receives only medication names and relevant dietary restrictions from your profile — not your full medical history, appointment notes, or unrelated documents. This reduces exposure and protects your privacy.

What the AI doesn't do

It does not diagnose conditions

The AI companion is not a diagnostic tool. It provides information and insights, but only your doctor can diagnose medical conditions.

It does not recommend treatments

The AI does not prescribe medications or recommend specific treatments. Always consult your healthcare provider before making medical decisions.

It does not replace your doctor

The AI companion is designed to help you understand and manage your health journey, not replace professional medical advice. Always work with your care team.

It does not handle emergencies

If you're experiencing a medical emergency, call 000 (Australia) or your local emergency number immediately. Do not use the AI companion for urgent medical situations.

Your data is not used to train AI models

Azure OpenAI API terms prohibit using your data to train or improve their models. This is contractually enforced.

When you chat with your AI companion, the conversation data is:

  • ✓Sent to Azure OpenAI to generate a response
  • ✓Stored in your RareJourney account (encrypted) so you can review chat history
  • ✓Deleted from Azure OpenAI's systems within 30 days to comply with abuse monitoring policies
  • ✗Never used to train or fine-tune Azure's AI models
  • ✗Never shared with other Azure customers or Microsoft products

Data residency & processing location

All AI processing happens in Australia East (Azure region) to ensure data residency and compliance with Australian privacy laws.

•Your data does not leave Australia during AI processing
•Azure OpenAI Service operates under Microsoft's Australian data residency commitments
•This meets the requirements of the Australian Privacy Principles (APP 8)

Understanding AI limitations

AI can make mistakes

While our AI companion is designed to be helpful and accurate, it's important to understand that AI systems can occasionally:

  • •Misinterpret complex medical information
  • •Generate plausible-sounding but incorrect information ("hallucinations")
  • •Miss nuances that a human healthcare provider would catch

Always verify important medical information with your healthcare provider. The AI is a tool to support your understanding, not a replacement for professional medical judgment.

Questions about AI transparency?

We're committed to helping you understand how our AI companion works. If you have questions or feedback, we'd love to hear from you.