AI Use Policy

Effective: 6 May 2026

Myndura is operated by Mynda Tech Solutions ("we", "us", or "our"). This AI Use Policy explains how the AI features in Myndura work, what they can and cannot do, and how we keep them safe and accurate.

If you have questions about this policy, contact us at support@myndura.com.

1. What the Myndura Guide Is

The Myndura Guide is an in-app conversational tutor that helps you learn psychiatry. It can:

  • Explain concepts that appear in a lab you are working through.
  • Answer cross-cutting clinical questions (e.g. "how does this differ from condition X?", "what does DSM-5-TR say about Y?").
  • Generate quick quizzes and summaries on demand.
  • Suggest follow-up questions to deepen your understanding.

The Myndura Guide is an educational aid. It is not a clinical advisor for real patients, not a diagnostic tool, and not a substitute for supervision, authoritative references, or your own clinical judgement.

2. What the Myndura Guide Is Not

The Myndura Guide is not:

  • A source of clinical decisions for real patients.
  • A prescribing reference for medication selection or dosing.
  • A risk-assessment instrument.
  • A simulated patient or role-play partner for clinical scenarios.
  • A replacement for authoritative sources (DSM-5-TR, ICD-11, NICE, BNF, FDA prescribing information, UpToDate, your local guidelines, your supervisor).

For any decision that affects a real patient, consult an authoritative source and your supervising clinician.

3. How It Works

The Myndura Guide is powered by large language models from Anthropic (Claude) and Google (Gemini). When you send a message:

  • Your message and the lab context (current step, author notes, glossary, references) are sent to the model provider.
  • The provider returns a response, which is streamed back to your screen.
  • We do not use your conversations to train AI models. The provider's data-use terms apply to API traffic and prohibit using your messages for model training.

The specific model that handles your message depends on your subscription tier and our model policy at the time of the request. We may switch between providers and models without notice as we tune for quality and cost.

4. Limitations and Accuracy

Large language models can make mistakes. The Myndura Guide can:

  • Confabulate specifics. Drug half-lives, prevalence percentages, dose thresholds, NNT/NNH values, year and author of studies, sample sizes, and specific DSM or ICD criterion counts are all categories where AI models occasionally produce confident but incorrect numbers. We instruct the Myndura Guide to flag uncertainty and point you to a primary source for these — but the safeguard is not perfect.
  • Reflect outdated training data. A model's training data has a cutoff date. Recent guideline updates (e.g. a NICE refresh, an FDA labelling change) may not be reflected.
  • Misrepresent disagreement. Where evidence is genuinely mixed or where DSM-5-TR and ICD-11 diverge, the model may oversimplify into a false consensus.
  • Hallucinate citations. AI models can produce plausible but fabricated PMIDs, DOIs, author names, or study findings. Always check a primary source before citing a paper the Guide names.

Always verify clinical claims against an authoritative source before acting on them.

5. What We Do To Keep It Useful

  • Curated lab content as ground truth. Each lab is authored and reviewed by clinical experts. The Myndura Guide is instructed to defer to the lab's content, references, and author notes when its training data conflicts with what the lab states.
  • Calibrated language. The Guide is instructed to distinguish "established in DSM-5-TR" from "general clinical understanding" from "I'd verify this in BNF / UpToDate" — so you can tell at a glance how much weight to put on a claim.
  • Source precedence. When you ask about specific numbers (dosing, half-lives, prevalence, study results), the Guide is instructed to point you to BNF, FDA prescribing information, UpToDate, or the primary literature rather than guessing.
  • Safety guardrails. The Guide will not simulate being a patient, role-play clinical scenarios, give direct clinical advice for real cases, or state specific medication dosages as recommendations. For safety-critical topics (suicide risk assessment, crisis management, paediatric or perinatal medication choices), it uses pre-approved educational frameworks only.
  • Your reports. Every Guide response has a flag icon. Tap it if the response is wrong, misleading, harmful, or off-topic. Reports go to our content team and feed into prompt and content improvements.

6. What You Should Do

  • Treat the Myndura Guide as a study aid, not a source of clinical truth. Use it to learn, not to decide.
  • Verify before acting. For any specific number, dose, criterion, or study claim, check the primary source.
  • Use the flag icon. If the Guide gives a response that is wrong or unsafe, flag it. We rely on these reports.
  • Do not enter real patient identifiers. Use de-identified scenarios only. See our Privacy Policy for what data we store from your conversations.

7. Data Handling

Your messages to the Myndura Guide are sent to the model provider (Anthropic or Google) over an encrypted connection. We do not use your conversations to train AI models, and the provider's API terms prohibit them from using your data for training.

We retain Guide conversations for usage analytics, abuse prevention, content improvement, and your own review history. See our Privacy Policy for full retention details, your rights of access and deletion, and how to request that data be removed.

8. Reporting Issues

If you encounter a Myndura Guide response that is wrong, misleading, or unsafe:

  1. Tap the flag icon next to the response inside the app, or
  2. Email support@myndura.com with a description of what happened.

For broader concerns about Myndura's AI features, contact us at support@myndura.com.

9. Changes to This Policy

We may update this policy from time to time as our AI features evolve. Material changes will be communicated in-app and via email. The "Effective" date at the top of this page reflects the most recent revision.

Mynda Tech Solutions, May 2026.