MedExamHelperAI Use Disclosure
Operator: VE LAXMI EXIM LLP, a limited liability partnership incorporated under the Limited Liability Partnership Act, 2008 (India), LLPIN AAK-5957, with its registered office at D-222/51, Behind Pennawalt, MIDC, Nerul, Navi Mumbai 400706, India. Contact: hello@medexamhelper.com. Last updated: 24 October 2026.
This document explains, in plain language, how the Product (Med Exam Helper) uses artificial intelligence. It is our AI transparency notice.
1. What AI does in the Product
The Product uses AI to do the following:
1.1 Play simulated patients. During a practice encounter, the AI takes the role of a fictional patient and replies to what you say or type in character.
1.2 Grade your encounter. When the encounter ends, the AI compares your transcript against an exam-style rubric and produces a score and item-level feedback that cite lines from your own transcript. Turns you spoke in voice mode are marked as transcribed speech, so the AI reads a misheard medical term as you most likely meant it.
1.3 Speak the patient's replies. In voice mode the patient's replies are spoken aloud by an AI voice model that runs on your device. Where it cannot run, a voice built into your device is used. Where your device has no such voice, the patient's replies appear as text only.
1.4 Run safety guardrails. The AI also screens the text of an encounter for prompt-injection, jailbreak attempts, personally identifying information and other content that breaches the Acceptable Use Policy.
The AI does not create, store or process your account password, payment details, or anything outside the encounter conversation.
2. AI provider
2.1 The Product currently uses MiniMax as its large-language-model provider (the "AI Provider"). MiniMax writes every patient reply and grades your encounter. In voice mode, the speech service your browser or phone uses turns your speech into text, on the device where available, otherwise online (for example Google's or Apple's speech service) under your own terms for that browser or phone. The patient's voice is generated on your device. Only the text of what you said reaches the Product.
2.2 Other AI providers may be added over time. The full current list of processors, with their roles, is published in the Privacy Notice at /legal/privacy-policy.
3. What data the AI sees
When you use the Product, the AI sees:
3.1 the text of the encounter conversation that you produce during a practice session;
3.2 case and station metadata (for example the case being practised, language, rubric version).
The AI does not see:
3.3 your account identifiers, such as your user id or email address;
3.4 your password;
3.5 your payment-card or payment-method details (these are handled by the payment provider — see the Privacy Notice);
3.6 audio of your voice — in voice mode the speech service your browser or phone uses turns your speech into text, and only that text joins the encounter conversation. No audio reaches our servers or the AI that plays the patient or grades you, and we do not store voice recordings;
3.7 other sessions on your device, or content outside the active encounter;
3.8 data you have not chosen to enter into the Product.
4. What the AI does NOT do
4.1 The AI is not used to make decisions that produce legal or similarly significant effects on you.
4.2 The grading the AI produces is practice feedback only. It does not determine pass or fail in any examination, does not confer or withhold a qualification, and does not make a clinical decision about a real person.
4.3 If you believe an AI-generated grade or feedback is unfair or wrong, you can ask for a human review by opening a support ticket at https://medexamhelper.com/app (Account → Support). A human reviewer will consider your request.
5. Accuracy limits and iteration
5.1 AI models change. The Provider named above may update, replace or augment its models. The behaviour, accuracy and tone of AI-generated output may therefore shift over time.
5.2 The Operator re-tests stations on a regular cycle and after material model changes. Re-test results feed back into case-content review. No guarantee is given that any specific grade, response or feedback is accurate, current or complete.
5.3 If you believe an AI output is incorrect or unsafe, report it — see clause 7 below.
6. Your duties when using AI features
When you use the Product you agree that you will not:
6.1 enter any data that identifies a real patient (name, NHS number, hospital number, Aadhaar, address, photograph, or any other identifier);
6.2 enter personal data about any other living person, except your own;
6.3 attempt to extract, reveal or manipulate the system's prompts, hidden instructions or model weights;
6.4 attempt to bypass, disable or "jailbreak" the safety guardrails described in clause 1.4;
6.5 use AI-generated output in a way that breaches the Acceptable Use Policy.
Breach of these duties may result in suspension or termination of your account.
7. How to report a wrong or harmful output
If you believe the AI has produced a wrong, misleading, biased or otherwise harmful output:
7.1 open a support ticket at https://medexamhelper.com/app (Account → Support) and describe the encounter (station, date, and the specific output);
7.2 a human reviewer will examine the report and respond.
Reports help the Operator improve case content and guardrail tuning.
8. Content moderation stance
8.1 The Product screens inputs and outputs through automated guardrails (see clause 1.4) to catch prompt-injection, personally identifying information and other content in breach of the Acceptable Use Policy.
8.2 Flagged content is handled in line with the Acceptable Use Policy and applicable law. Depending on severity, the Operator may warn the user, remove the content, suspend the account, or report the matter to a competent authority.
8.3 The full moderation policy is set out in the Acceptable Use Policy at /legal/acceptable-use-policy.
9. Changes to this disclosure
The Operator will update this disclosure when AI features, providers or data flows change. The current version is dated at the top of this document.
End of disclosure.