Healthcare Practice Privacy Policies: Use case and AI Confidentiality
A practice that holds health information needs a clearly expressed and up-to-date privacy policy, and someone has to refresh it every year. At Kestrel Lane the practice manager pastes last year’s text, a list of the practice’s systems and vendors and her own drafting notes into ChatGPT or Claude and asks for a plain-English rewrite with a paragraph on AI. The published words are meant to be public. The notes around them are not: how records are held and backed up, which vendor sends data to the United States, what is unfinished and who signs off. This page marks each item in a draft privacy policy that the Privacy Act and NSW health records law bear on, and shows how the same draft is worked on Ermos inside the practice.
The practice privacy policy
Written or reviewed by a practice manager or principal, usually once a year or when a new system such as an AI scribe is introduced. The Privacy Act asks an entity to make its privacy policy available free of charge and in an appropriate form, which usually means the website.
Privacy policy
Kestrel Lane Medical Practice · DRAFT 7 for annual review · Wednesday 18 November 2026
1. About this policy. Kestrel Lane Medical Practice, 14 Kestrel Lane, Wollstonecraft NSW 2065. Our privacy contact is Helen Tran, Practice Manager, on (02) 5550 0142.1 This policy explains how we handle your personal information, including your health information.
2. What we collect. We collect and hold your name, address, date of birth, Medicare, DVA and private health fund numbers, emergency contact, medical history, medicines, allergies, test results, images and the notes of your consultations.2
3. How we collect and hold it. Records are held electronically in Hollowell Clinical (Hollowell Practice Systems Pty Ltd), hosted in Sydney, and Carrow IT Services can reach them remotely to support the system. [Drafting note, Helen: the nightly backup goes to a second Hollowell site and Carrow says “an Australian region” but has not put it in writing. Paper records from before 2019 are in off-site storage and the destruction register is still not done.]3
4. Why we collect it and who we share it with. To provide your care, run the practice and meet our legal duties. We share information with specialists, pathology and imaging providers, hospitals, Medicare and My Health Record when you agree or the law allows, and with our insurer if a claim arises.4
5. AI scribe. Some of our doctors use an AI scribe that listens to a consultation and drafts the note. It is a trial with two doctors. We ask for your consent at the start of each consultation and you can say no at any time.5
6. Overseas disclosure. We are likely to disclose some information to a recipient in the United States, which runs the language model behind the AI scribe. [Drafting note, Helen: Nightjar’s draft contract lets it use de-identified text to improve the service unless we opt out in writing. Dr Voss wants a line saying we never use your information to train AI. Do not publish it until Nightjar confirms in writing. Countries to be confirmed.]6
7. Access, correction and how long we keep records. Ask for your records in writing and we will respond within 45 days. You can ask us to correct them. We keep adult records for seven years after your last visit and children’s records until they turn 25.7
8. Complaints. Tell Helen Tran in writing and we will reply in writing. If you are not satisfied you can complain to the Office of the Australian Information Commissioner or the NSW Health Care Complaints Commission.8
9. Automated decisions. [Drafting note, Helen: from 10 December the policy has to describe certain automated decisions. Our online booking tool hides same-day slots from anyone who answers yes to the urgent-care questions and tells them to call. Is that a decision? Ask our adviser before 26 November.]9
DRAFT 7, prepared by Helen Tran. For approval by Dr Helena Voss at the practice meeting on Thursday 26 November, then on the website and at reception by Friday 4 December. Last published version: 20 November 2025.10
An invented policy draft. The practice, its software vendor and its staff are fictional. Hover or tap any highlighted item for the analysis; the full list is below.
Privacy policy update drafter
An Agent+ that prepares the yearly privacy policy review. It compares the current published policy with the practice's change notes, such as a new system, and drafts updated wording in the practice's style. Drafting notes and unconfirmed statements are pulled into a separate list, away from the public text.
What it does
- Reads the current policy and the practice's change notes
- Drafts updated wording in the practice's own style
- Lists open notes and statements still unconfirmed
What stays with your people. The principal and practice manager confirm every statement, and take legal advice where needed, before publishing. The agent drafts; it does not publish.
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
- 1The practice name and privacy contact
The practice’s address and a named contact. It is already public and a published policy is meant to carry it. APP 5.2(a) and HPP 4(1)(a) ask that the individual is made aware of the organisation’s identity and how to contact it. Privacy Act guide.
Frontier Models (ChatGPT, Claude) Already public once the policy is published; nothing private here Ermos Edge Not about where data goes Ermos Dominion Not about where data goes Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 2The kinds of information collected
A generic list of the kinds of personal information the practice collects and holds. APP 1.4(a) says the privacy policy must contain it. No patient is named. A practice that provides a health service and holds health information is not a small business operator (Privacy Act s 6D(4)(b)), so its turnover does not take it outside the Act. Privacy Act guide.
Frontier Models (ChatGPT, Claude) Generic categories with no patient named; harmless to send Ermos Edge Not about where data goes Ermos Dominion Not about where data goes Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 3How records are held, the vendors and the unfinished work
APP 1.4(b) asks the policy to say how the entity collects and holds personal information, and APP 11.1 asks for reasonable steps to protect it. This item goes further than a public policy would: it names the software vendor, the IT contractor with remote access, an unconfirmed backup location, off-site paper records and a missing destruction register. HPP 5(1)(d) asks that everything reasonably within the organisation’s power is done to prevent unauthorised use or disclosure when information is given to a person providing a service to it. APP 1.3 asks for a clearly expressed and up-to-date policy, so the practice rather than an AI tool has to confirm each statement. Privacy Act guide.
Frontier Models (ChatGPT, Claude) The records set-up, a vendor with remote access, an unconfirmed backup and unfinished work transmitted Ermos Edge Stays on your instance, but a person must still confirm where the backups are Ermos Dominion Stays on the unit, but a person must still confirm where the backups are Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 4Purposes and who information is shared with
APP 1.4(c) asks for the purposes for which the entity collects, holds, uses and discloses personal information. The wording is generic. Under APP 6.1 and HPP 11(1) a disclosure for another purpose needs consent or an exception such as a directly related purpose the patient would reasonably expect. Privacy Act guide.
Frontier Models (ChatGPT, Claude) Generic wording; harmless to send, but a person must check it matches what the practice does Ermos Edge Not about where data goes Ermos Dominion Not about where data goes Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 5The AI scribe paragraph for patients
The OAIC says organisations should update their privacy policies and notifications with clear information about their use of AI, and Ahpra asks practitioners to tell patients about AI and obtain informed consent. The paragraph is written for patients and names no one. Whether it matches what the doctors do is for the practice to check. Patient data guide.
Frontier Models (ChatGPT, Claude) A public paragraph that names no one; harmless to send Ermos Edge Not about where data goes Ermos Dominion Not about where data goes Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 6The overseas recipient and the training question
APP 1.4(f) and (g) ask whether the entity is likely to disclose personal information to overseas recipients and, if practicable, the countries. APP 8.1 asks for reasonable steps before the disclosure to ensure the overseas recipient does not breach the APPs (other than APP 1), subject to the exceptions in APP 8.2, and HPP 14 deals with transfers of health information outside NSW. The note beside it records a vendor sub-processor, a contract term the practice has not accepted and a promise the doctor wants to make but cannot yet keep. A confident AI rewrite could publish the promise. Privacy Act guide.
Frontier Models (ChatGPT, Claude) A named vendor, a United States sub-processor, a contract gap and a pending promise transmitted Ermos Edge Stays on your instance, but what the vendor does with the data is for the practice to settle with it Ermos Dominion Stays on the unit, but what the vendor does with the data is for the practice to settle with it Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 7Access, correction and retention periods
These match the NSW Act: a private sector person must respond to a written request for access within 45 days (s 26(1)(a) and s 27(1)) and a health service provider keeps adult records for 7 years from the last occasion of service and children’s records until the age of 25 (s 25(1)). APP 1.4(d) asks how an individual may access and seek correction of information. For an organisation, APP 12.4(a)(ii) asks for a response within a reasonable period. A rewrite can change a period without anyone noticing. Privacy Act guide.
Frontier Models (ChatGPT, Claude) A public rule with no patient data; harmless to send, but the draft can restate a period wrongly Ermos Edge Stays on your instance, but a person must check each period against the Act Ermos Dominion Stays on the unit, but a person must check each period against the Act Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 8How to complain
APP 1.4(e) asks the policy to say how an individual may complain about a breach of the APPs and how the entity will deal with the complaint. The wording is generic and public. The HCCC deals with complaints about health service providers in NSW.
Frontier Models (ChatGPT, Claude) Generic public wording; harmless to send Ermos Edge Not about where data goes Ermos Dominion Not about where data goes Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 9The automated decisions note
Privacy and Other Legislation Amendment Act 2024, Schedule 1 Part 15, inserts APP 1.7 to 1.9 from 10 December 2026. The privacy policy must then say what kinds of personal information are used by computer programs that make, or do a thing substantially and directly related to making, a decision that could reasonably be expected to significantly affect an individual’s rights or interests, and the kinds of those decisions. Whether a booking rule is such a decision is a question for the practice’s adviser. The note shows the practice has not settled it. Privacy Act guide.
Frontier Models (ChatGPT, Claude) An open question about the practice’s booking tool; low sensitivity alone Ermos Edge Stays on your instance, but only the practice’s adviser can say whether it is a decision Ermos Dominion Stays on the unit, but only the practice’s adviser can say whether it is a decision Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case - 10Approval, dates and the previous version
Who prepared the draft, who approves it and when it goes live. APP 1.5 asks for the policy to be available free of charge and in an appropriate form, which the Act’s note says is usually the entity’s website. The internal timetable and the date of the last version are not public.
Frontier Models (ChatGPT, Claude) Names the practice manager and the doctor with the internal timetable; low risk alone Ermos Edge Not about where data goes Ermos Dominion Not about where data goes Use case Privacy policy update drafter. Drafts an updated privacy policy from the current text and the practice's change notes, and lists open drafting notes and unconfirmed statements for review. See the use case
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 Edge | Ermos Dominion | |
|---|---|---|---|
| Where the document is processed | Provider servers, generally offshore on personal plans | Your instance in an Australian data centre; answers generated in Australia | The unit in your office |
| Who can read it | Provider staff and systems, under policy | Your staff, within the boundaries you set. The named Australian provider processes only permitted excerpts, under contract, and keeps nothing | Your staff only |
| Is it kept | On the provider’s terms | Kept on your instance; the provider is barred from retaining it | Stays on the unit |
| Can it train a model | On consumer tiers, yes | No, by contract | Only your own unit learns from it, on your hardware |
Part of Use cases in healthcare practices. See how this works in practice: Practice policies and forms on Ermos.
Common questions
Do ChatGPT and Claude use the information in a medical practice privacy policy draft?
Is it safe to paste a draft practice privacy policy into ChatGPT or Claude?
Can ChatGPT or Claude be trusted to draft a medical practice privacy policy?
How can a medical practice use AI for a privacy policy without sending it to ChatGPT or Claude?
Does a medical practice need a written privacy policy?
What does a privacy policy have to say about overseas disclosure?
What changes in privacy policies from 10 December 2026?
See Dominion answering from your own documents.
A demonstration takes 30 minutes, uses no client data, and comes with no obligation.
Book a demo