Healthcare GP Registrar Case Review Forms: Use case and AI Confidentiality

A GP who supervises a registrar writes up each teaching session: the cases discussed, what went well, what needs work and what was agreed. To turn her rough notes into feedback and a learning plan, she pastes the form into ChatGPT or Claude. The form names three patients with ages and conditions, records a clinical error, includes a young person’s confidential request, quotes what the registrar said about her own health, rates her performance and notes an agreement about using AI. This page marks each item that the Privacy Act, NSW health records law and Good medical practice bear on, and shows how the same form is worked on Ermos inside the practice.

The GP registrar case review form

Completed by a supervising GP after a teaching session with a GP registrar. Practices and training providers use different formats, and this is the practice’s own form rather than a College template. Good medical practice section 12 describes teaching, supervising and assessing as part of good medical practice.

Registrar case review (supervisor’s form)
Kestrel Lane Medical Practice · Thursday 12 November 2026, 12.30pm to 1.15pm, protected teaching time

Registrar and supervisor
Case 1
Case 2
Case 3
Supervisor’s assessment
Registrar’s wellbeing
Agreed actions
Use of AI
Basis for the cases
Next review and sign-off

An invented form. The registrar and the patients are fictional; the practice, Dr Voss and the dates sit in the same invented timeline as the other healthcare documents, and the registration number is a placeholder. Hover or tap any highlighted item for the analysis; the full list is below.

Registrar case review form drafter

An Agent+ that drafts the registrar case review form from the supervising GP's notes after a teaching session. It uses your practice's own form, sets out each case, the learning points and the follow-up, and flags where a case touches your policy on AI tools.

What it does

  1. Reads the supervisor's notes from the session
  2. Drafts each case in your review form
  3. Flags items against your policy text

What stays with your people. The supervising GP edits, completes and keeps the form. Any assessment of the registrar stays with the supervisor.

An Agent+ is built once and anyone in the practice can run it. It works from your own documents and runs on Ermos Dominion, the AI unit in your office, or on Ermos Edge, your own instance in an Australian data centre.

What leaves the practice

  1. 1
    The registrar, the supervisor and the registration number

    Names two doctors, a start date and a registration number. Personal information about the registrar, tied to a document that goes on to assess her. Registration numbers are on the public register, but here the number sits beside an assessment. The number shown is a placeholder.

  2. 2
    A named patient, a diagnosis and a wound

    A name, an age, a chronic condition and where the wound is: health information (Privacy Act s 6FA). Using patient records for teaching is a secondary use. APP 6.2(a)(i) allows a use for a secondary purpose directly related to the primary purpose that the person would reasonably expect, where the information is sensitive. HPP 10(1)(e) in NSW allows a use that is reasonably necessary for training employees or people working with the organisation if, for example, the purpose cannot be served by non-identifying information and consent is impracticable, or reasonable steps are taken to de-identify, and the information is not published in a generally available publication. The same paragraph also requires the use to be in accordance with any guidelines the Privacy Commissioner has issued for it, and the Commissioner has issued statutory guidelines on training: the Information and Privacy Commission says they require the people being trained, and anyone who will access the health information during the training, to sign an agreement that they are aware of the Health Privacy Principles and will comply with them. HPP 10 deals with use by the organisation; whether sending a case to an outside AI tool is a use or a disclosure is a separate question, because HPP 11 limits disclosure. Patient data guide.

  3. 3
    A clinical error and a medicine history

    A patient who is named, a clinical mistake and what she took. Health information, and an account of a registrar’s error written for learning. Good medical practice section 12 covers teaching, supervising and assessing, and a supervisor remains responsible for the feedback. The medicine and the sequence of events are the details that make the case identifiable even if the name is removed. Patient data guide.

  4. 4
    A 16-year-old’s confidential request and her mother’s name

    A young person’s request for contraception, her wish that her mother not be told and the mother’s name. Health information about the young person, and a family detail that identifies her. Good medical practice 4.4.1 asks doctors to treat information about patients as confidential. When a young person can consent to treatment and what a parent may be told are questions for the practice’s clinical and legal advice, and this page does not answer them; a draft written by an AI tool could state them wrongly. Patient data guide.

  5. 5
    The supervisor’s candid assessment

    An opinion about a named person’s performance. If the registrar is employed by the practice, the Privacy Act exempts an act or practice of an employer directly related to the employment relationship and an employee record it holds (s 7B(3)), but it applies only if the practice is the registrar’s employer, and the exemption does not reach the patient information in the same form. Good medical practice 12.1 recognises a power imbalance between supervisors and those they supervise. A draft from an AI tool may change the tone or the rating, and the supervisor remains responsible for the assessment.

  6. 6
    The registrar’s own health and her father’s

    Sleep problems, a relative’s hospital admission and an agreement to seek care. Information about a person’s health, including an illness, is health information (s 6FA(a)(i)), and sensitive information (s 6(1)). Good medical practice has a section on ensuring doctors’ health. The employee record exemption in s 7B(3) covers acts directly related to the employment relationship; whether it reaches a transfer to an outside AI tool is a question for the practice’s legal adviser. Privacy Act guide.

  7. 7
    The agreed actions and dates

    A learning plan with dates. It names the registrar and the supervisor and no patient beyond the one case. Low risk alone.

  8. 8
    The agreement about using AI

    Records the practice’s own rule and that none is yet written. Ahpra asks practitioners to be aware of the governance arrangements their practice sets up for AI. The accreditation report of 4 November already found no written scribe policy. The rule is the one the supervisor breaks if she pastes this form into an outside tool. Patient data guide.

  9. 9
    How the cases were chosen and whether patients knew

    States the practice’s basis for using the cases and says patients were not asked. APP 5 and HPP 4(1)(c) and (d) ask an organisation to take reasonable steps to make a person aware, at or before collection or as soon as practicable, of the purposes of collection and the persons it usually discloses to. Whether teaching is among the purposes the practice tells patients about is a fact about the practice’s privacy notice. HPP 10(1)(e) also turns on de-identification, impracticability of consent and the Privacy Commissioner’s statutory guidelines on training. Privacy Act guide.

  10. 10
    The next review and the signatures

    Names the two signatories and the date of the next session. It shows who is accountable and when it will be revisited.

What happens in Frontier Models (ChatGPT, Claude)

  • The whole document is transmitted to the provider’s servers. On personal plans the user cannot choose where it is stored or processed, and for the major providers that has generally meant outside Australia, under that country’s legal process. Some business plans offer storage in Australia, but the provider still receives the content.
  • On consumer tiers, prompts can be retained and used to train future models.
  • The provider’s staff and systems can reach the content under the provider’s own policies.
  • Once sent, the practice no longer controls it. Retention and deletion happen on the provider’s terms.

The same document on Ermos

Ermos Dominion

The document is drafted on a unit in your office, from your own templates and the patient record. Every marked item stays on the unit, which keeps working with the internet switched off. Nothing in it is transmitted to an AI provider, so there is no disclosure to analyse: the confidentiality work is the same as for your file server.

Ermos Edge

Your documents, the index and your access boundaries stay on your own instance in an Australian data centre. When a staff member asks for this draft, only their question and the excerpts they are permitted to see go to a named Australian provider, who is contractually barred from keeping the content or training any model on it. Nothing is processed outside Australia.

Frontier Models (ChatGPT, Claude)Ermos EdgeErmos Dominion
Where the document is processedProvider servers, generally offshore on personal plansYour instance in an Australian data centre; answers generated in AustraliaThe unit in your office
Who can read itProvider staff and systems, under policyYour staff, within the boundaries you set. The named Australian provider processes only permitted excerpts, under contract, and keeps nothingYour staff only
Is it keptOn the provider’s termsKept on your instance; the provider is barred from retaining itStays on the unit
Can it train a modelOn consumer tiers, yesNo, by contractOnly your own unit learns from it, on your hardware

Part of Use cases in healthcare practices. See how this works in practice: Clinical documentation on Ermos.

Common questions

Do ChatGPT and Claude use the information in a GP registrar case review form?
Yes, a pasted case review form is sent to the AI company's servers and processed there to produce a reply. As at October 2026, whether a conversation on a personal plan (ChatGPT Free, Plus or Pro; Claude Free, Pro or Max) can be used to train the models depends on a setting each user controls. Business plans (ChatGPT Business and Enterprise, Claude Team and Enterprise) and the developer APIs sit under commercial terms, and OpenAI and Anthropic each say they do not train on that content by default (OpenAI, Anthropic). On either product, the thumbs up and thumbs down buttons can send a whole conversation back for training. This form names three patients with their ages and conditions, records a 16-year-old's confidential request for contraception beside her mother's name, and quotes the registrar on her sleep and her father's hospital admission.
Is it safe to paste a GP registrar case review form into ChatGPT or Claude?
No, because the form holds patient information and an assessment of a named doctor in the same paste. The Office of the Australian Information Commissioner (OAIC) recommends in its guidance on commercially available AI products that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools, because of the significant and complex privacy risks. Taking out the names leaves a woman of 29 with eyelid swelling on 4 November who had started naproxen the week before, which is the detail that makes Case 2 identifiable. Removing names rarely makes a document anonymous. Information stays personal information while the person is reasonably identifiable, and the OAIC warns that information can be at risk of re-identification even when it has been de-identified or anonymised. HPP 10(1)(e) in NSW covers use for training, and it turns on de-identification and on the Privacy Commissioner's guidelines, which the Information and Privacy Commission says require trainees and anyone who will access the information to sign an agreement to comply with the Health Privacy Principles. Whether sending a case to an outside AI tool is a use or a disclosure is a separate question, because HPP 11 limits disclosure.
Can ChatGPT or Claude be trusted to draft a GP registrar case review form?
Not without the supervisor checking every line. Language models can produce fluent text with details that are not in the source, wrong figures, wrong dates or references that look real but are not. On this form the most harmful errors would be a changed account of the medicines in Case 2, a wrong statement about what the law lets a mother be told about a 16-year-old, a softened or hardened rating in the assessment, or a wrong date for the actions due on 16, 23 and 27 November. Dr Voss remains responsible for the feedback, so she should check each item against her own notes and the record.
How can a GP practice use AI for a GP registrar case review form without sending it to ChatGPT or Claude?
Ermos is built so the document stays under your control. On the Dominion AI unit it is drafted inside your office and none of it is sent to an AI provider. On Ermos Edge your documents and index stay on your own instance in an Australian data centre, and only the question and the excerpts the user is permitted to see go to a named Australian provider that is contractually barred from keeping them or training any model on them. The feedback and learning plan are drafted from the supervisor's notes and the practice's own form, and Dr Voss checks every line and signs it. That fits the rule she gave Dr Mirchandani on 12 November, that no identifying details go to ChatGPT.
Can a practice use patient records to teach a GP registrar?
In NSW, HPP 10(1)(e) allows use of health information for a secondary purpose reasonably necessary for training employees or people working with the organisation, if the purpose cannot be served by non-identifying information and consent is impracticable, or reasonable steps are taken to de-identify it, and it is not published in a generally available publication. HPP 10(1)(b) also allows a directly related use the person would reasonably expect. The Privacy Commissioner's statutory guidelines on training also apply to HPP 10(1)(e): the Information and Privacy Commission says they require the people being trained, and anyone who will access the information during the training, to sign an agreement that they are aware of the Health Privacy Principles and will comply with them.
Does the Privacy Act cover a supervisor's notes about a registrar?
Section 7B(3) of the Privacy Act exempts an act or practice of an employer that is directly related to a current or former employment relationship and an employee record the employer holds. Whether it applies depends on who employs the registrar. The exemption does not change the protection of patient information in the same document, and whether it covers sending the notes to an outside tool is a legal question.
What does Good medical practice say about teaching and supervising?
Section 12 of Good medical practice covers teaching, supervising and assessing. It says it is part of good medical practice to contribute to these activities and to provide support, assessment, feedback and supervision for colleagues, doctors in training and students. It also asks doctors to recognise the power imbalance between supervisors and those they supervise and not to take advantage of it.

See Dominion answering from your own documents.

A demonstration takes 30 minutes, uses no client data, and comes with no obligation.

Book a demo