Use cases and AI confidentiality in healthcare: what leaves when patient records go into ChatGPT or Claude.

A medical practice holds the most sensitive records most people will ever have written about them: diagnoses, medications, mental health, family history and what was said in the room. Under the Privacy Act health information is sensitive information, and a private business that provides a health service and holds health information is covered whatever its turnover. In NSW the Health Records and Information Privacy Act 2002 adds its own principles, and Good medical practice asks doctors to keep records accurate and secure and information confidential. This series takes 52 documents a general practice with allied health produces every week, marks every item those obligations reach, and shows where each one goes when the document is pasted into ChatGPT or Claude. Each page ends with the same document on Ermos, inside the practice.

The documents, one at a time

LetterThe GP referral letter

Identity, symptoms, history, medications, results and a bereavement. A patient’s whole clinical picture, in one page.

See what leaves →
NotesThe GP consultation note

Symptoms, alcohol, a family cancer history, an abnormal result and an apology. What the patient said before it became a plan.

See what leaves →
FormThe new patient intake form

Identity, Medicare, mental health history, drug use, an emergency contact and a choice about AI. The first page of the record.

See what leaves →
EmailThe internal abnormal pathology result email

A patient, the values, a suspected cancer and how to call without telling the household. Written in a hurry before leave.

See what leaves →
RecordThe GP mental health treatment plan

Scores, a diagnosis, a risk assessment, substance use and who must not be told. The most intimate page in the record.

See what leaves →
SpreadsheetThe billing and recall spreadsheet

Names against birth dates, mobiles, item numbers, condition codes and notes. A whole patient list, one row each.

See what leaves →
TranscriptThe AI scribe consultation transcript

The consent line, the symptoms, a bereavement, drinking, a family dispute and another patient’s name. Every word as spoken.

See what leaves →
Report or memoThe medical report to an insurer

A claim number, a diagnosis, the treatment, prior depression, drinking, an opinion on work capacity and the patient’s finances.

See what leaves →
Text or chat messageThe appointment and recall SMS thread

A name and number, a doctor, a blood test, a family cancer history and a request to stop texting. A recall, in eight lines.

See what leaves →
LetterThe patient complaint response

The patient, a missed follow-up, the findings, the fixes, the indemnity notice and where to complain next. A bad week, in writing.

See what leaves →
SpreadsheetThe Medicare claims reconciliation spreadsheet

Names against item numbers, dates, fees, gaps and a column of who still owes. A clinic’s week of care, one row each.

See what leaves →
FormThe informed consent form for surgery

The procedure, the risks, the alternatives and the questions he asked. A patient’s fears and his wife’s illness, signed and dated.

See what leaves →
FormThe telehealth consent and identity check form

An identity check, a home address, an emergency contact, a referral, a gap fee and a privacy notice. Everything before the video call starts.

See what leaves →
LetterThe specialist letter to the referring GP

Test results, a diagnosis, new doses, a flight to Perth and a widow’s fear of hospital. What the specialist found and what happens next.

See what leaves →
NotesThe psychologist’s session notes

A risk assessment, panic attacks, drinking, a father who criticises and the therapist’s private impressions. Session three of six.

See what leaves →
RecordThe physiotherapy treatment plan and progress record

A knee replacement, diabetes, four sessions of measurements, a wedding, a wife with Parkinson’s and who pays for each visit.

See what leaves →
Report or memoThe workers compensation medical report

A shoulder injury, a claim number, an MRI, a certificate of capacity and a worker’s worry about his marriage. Written for an insurer.

See what leaves →
LetterThe letter responding to a request for health records

What is in her file, the page that is being withheld, a husband’s phone call, a fee and an unpaid account. The reply to a records request.

See what leaves →
Report or memoThe clinical incident report

A patient, a fracture, a blood thinner, a witness, an apology, a notification decision and an insurer notice. The week a clinic would rather keep inside.

See what leaves →
NotesThe clinician handover note

Four patients, a red-flag list, a fracture, an insurer claim, a power of attorney and three results to chase. Everything the cover clinician needs and the patients never meant to share.

See what leaves →
TranscriptThe after-hours mental health crisis call

A recording notice, a name, a bad night, the answers to a risk screen, a safety plan and consent to call a partner. The most sensitive eleven minutes in the record.

See what leaves →
Text or chat messageThe practice staff group chat

A patient moved because of a nosy manager, a caller who swore at reception, a pregnant patient’s result still missing and a photo of a form. All in one thread.

See what leaves →
SpreadsheetThe referral tracking spreadsheet

Initials, ages, referral types, referrers, urgency and days waiting. A small clinic’s waiting list, one row per person.

See what leaves →
LetterThe letter to a patient after an adverse outcome

The patient, a wound infection, a missed review, an apology, the review that follows and a note that the insurer has been told. A hard letter, in writing.

See what leaves →
Report or memoThe occupational therapy functional capacity report

A diagnosis, five assessment scores, a home with twelve steps, a husband, a sister, a job and a list of supports. The whole picture of a participant’s life.

See what leaves →
RecordThe post-discharge follow-up record

A pneumonia admission, a new blood thinner, a My Health Record and a wife with dementia who cannot be left alone. One patient’s first week home.

See what leaves →
FormThe GP chronic condition management plan

Diabetes, a heart investigation, a fear of hospital, three allied health referrals and a consent line. A whole patient, ready for a team.

See what leaves →
NotesThe home medicines review note

Twelve medicines, a falls risk, a Schedule 8 opioid, a lonely widower and a missed evening dose he does not want his daughter told about.

See what leaves →
Report or memoThe eligible data breach assessment

A misdirected result, a hepatitis B finding, a recipient who deleted it and a 30 day clock. The worst week of a practice manager’s year, written down.

See what leaves →
EmailThe email drafting a reply to a patient’s Google review

A one star review, a chest pain visit, a chart pasted into an AI tool and a reply that says far too much. A reputation problem that becomes a privacy breach.

See what leaves →
LetterThe letter responding to a regulator’s complaint

A chest pain visit, a complaint to the regulator, a clinical chronology, records enclosed and an insurer reading over the doctor’s shoulder.

See what leaves →
Report or memoThe NDIS allied health progress report

A diagnosis, a walking test, a fall at a bus stop, hours used and a plan review. A young woman’s independence, summarised for a funder.

See what leaves →
FormThe dental informed financial consent form

Three stages, a $7,200 estimate, a smoker with diabetes on warfarin, a finance application and a job lost in September.

See what leaves →
EmailThe employer’s email about a medical certificate

An HR manager, a certificate for a week off, a question about the diagnosis and a patient who has asked that his manager never know. A courtesy reply that can go wrong.

See what leaves →
LetterThe child immunisation history and catch-up letter

A four-year-old, a Scottish child health booklet, nothing on the national register, a chickenpox history and a father who wants the record.

See what leaves →
Report or memoThe general practice accreditation assessment report

Three indicators rated not met, a recall list facing the waiting room, a shared login, a destruction register that does not exist and an AI scribe with no policy.

See what leaves →
LetterThe reply to a request for a deceased patient’s records

A dead man’s heart failure, his daughter the executor, a son he did not want told, and a second request on the phone. A hard letter to get right.

See what leaves →
EmailThe notification of circumstances to an indemnity insurer

A wound infection, two nights in hospital, a review that was never booked and a policy that ends in February. The email a doctor sends before anyone has complained.

See what leaves →
Report or memoThe AI scribe vendor assessment

A vendor with no ARTG answer, audio in Australia and a model in the United States, a patient’s drinking still in the transcript, seven errors in forty notes and a nurse with a free chatbot.

See what leaves →
SpreadsheetThe clinical photo register

Nine patients, nine images: a diabetic heel, a child’s rash, a burn with a tattoo, three photos on personal phones and one on a WhatsApp group.

See what leaves →
FormThe GP registrar case review form

Three patients by name, a missed medicine history, a 16-year-old who does not want her mother told, a registrar who is not sleeping and a rule about AI the supervisor is about to break.

See what leaves →
Statement or scheduleThe practice privacy policy

A published policy that is meant to be public, wrapped in drafting notes about a backup nobody can locate, a United States sub-processor and a decision due on 10 December.

See what leaves →
SpreadsheetThe missed appointment log

Twelve entries in four weeks: a chest pain admission, a husband’s death, a mole result nobody has discussed and a voicemail left for a woman who asked for none.

See what leaves →
RecordThe nutrition care plan

A diabetes plan with an HbA1c, a BMI, a heart investigation, two or three glasses of wine most nights kept from a daughter and a consent line that never mentions AI.

See what leaves →
Report or memoThe psychological assessment report

A referral, an employer on a performance plan, a brother’s diagnosis, four test scores, a partner’s account and a diagnosis. The most complete picture of a client the clinic writes.

See what leaves →
SpreadsheetThe outcome measures tracking spreadsheet

Eleven clients, baseline and latest scores, a risk note, a young person, a wife’s illness and two clients labelled as not responding.

See what leaves →
RecordThe clinical supervision log

Four sessions, a client described by initials who is easy to identify, a full name, a risk discussion, a supervisee’s own confession and a supervisor’s verdict on her competence.

See what leaves →
LetterThe discharge from service letter

A back pain history, three missed sessions, a wife’s pain tablets, red-flag advice and a copy to the GP. The letter that ends a treatment relationship.

See what leaves →
SpreadsheetThe food diary analysis spreadsheet

Seven days of meals, daily weights, ferritin, cramps and dizziness, an eating disorder history and a mother who says she looks thin.

See what leaves →
RecordThe diabetic foot assessment record

A heel ulcer, a monofilament result, toe pressures, a SINBAD score, worn-out running shoes, a lost job and a patient who asked that nothing go to his sister.

See what leaves →
FormThe home visit risk assessment form

A new hip, a son in the granny flat, a rumour about a police callout, two dogs, a locked rifle safe, oxygen in the lounge and a code word.

See what leaves →
EmailThe insurer treatment approval request email

A claim number, a rear-end crash, neck disability scores, a 2023 neck visit, anxiety at an intersection and a sofa bed at his parents’ house.

See what leaves →

How the leakage happens

The mechanism is the same for every document. The whole text 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. On consumer tiers it can be retained and used to train future models. The provider’s staff and systems can reach it under the provider’s own policies. Once sent, the practice no longer controls it. Our on-premise vs cloud comparison sets out the four ways to run AI and where each one puts your documents.

The obligations in play

Each marked item on these pages links to the guide that covers its obligation:

The same documents on Ermos

Ermos Dominion drafts and searches on a unit in your office, so the documents never leave it. Ermos Edge keeps your documents on your own instance in an Australian data centre and sends only permitted excerpts, per question, to a named Australian provider. Every page in this series ends with both, item by item.

Common questions

Can healthcare practices use ChatGPT or Claude with patient information?
Not without meeting the practice's privacy duties, which still apply to every 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. Health information is sensitive information under the Privacy Act, and a business that provides a health service and holds health information is generally covered even if it is small. APP 6 limits using or disclosing it for another purpose unless the patient consents or an exception applies, and Ahpra says the practitioner remains responsible and must apply human judgment to any AI output.
Do ChatGPT and Claude train on what staff paste in?
It depends on the account each person is using. 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. For a medical practice the question matters because what staff paste is a patient's diagnosis, medications and what was said in the consulting room.
Do business plans such as ChatGPT Business or Claude Team solve the problem for healthcare practices?
Only in part. A business plan changes the default on training, but it does not change who receives the content. On a personal plan the user cannot choose where conversations are stored or processed. Some business plans let an organisation choose to store data at rest in Australia (OpenAI), which still leaves the content with the AI company. The OAIC's guidance also tells organisations to review carefully the terms and settings that will apply to their use of any AI product. For a practice, a referral or a mental health treatment plan would still arrive at that AI company in full, and Ahpra's guidance asks practitioners to know where patient data is located. Our comparisons with ChatGPT Enterprise and Claude weigh the trade-offs.
What is the safest way for a medical practice to use AI with patient documents?
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. For a healthcare practice, staying in control means patient records stay where the practice can protect them, and a doctor checks every draft that reaches the record.
What is AI data leakage in a healthcare practice?
Patient information leaving the practice's control through an AI tool. When a referral, a note or a transcript is pasted into ChatGPT or Claude it is transmitted to the provider's servers, can be retained on the provider's terms, can be used to train future models on consumer tiers, and can be reached by the provider's staff and systems. What leaks is a patient's health, often with their family and employer.
Does the Privacy Act apply to a small medical practice?
Generally yes. Section 6D(4)(b) of the Privacy Act takes a business out of the small business exemption if it provides a health service to another individual and holds health information other than in an employee record. The OAIC says the same: an organisation that provides a health service and holds health information is covered, even if it is a small business.
What do Ahpra and the RACGP say about AI in practice?
Ahpra says the practitioner remains responsible, must apply human judgment to AI output, should know where patient data is stored and whether it trains the model, and should obtain informed consent for AI scribes. The RACGP asks practices to get patient consent at the start of each consultation using a scribe and ideally to have the data stored in Australia.

See Dominion answering from your own documents.

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

Book a demo