Healthcare Specialist Letters to Referring GPs: Use case and AI Confidentiality

After the consultation the specialist writes back to the doctor who referred the patient: what was found, what will happen next and which medicines have changed. Specialists and their assistants paste the dictation or the draft into ChatGPT or Claude to tighten it, fix the layout or produce a plain-English copy for the patient. The letter carries test results, a diagnosis, doses, a family history and a note about a bereavement. This page marks every item in an invented reply from a cardiologist, names the duties each touches under the Privacy Act and the NSW health records law, and shows how the same letter is drafted on Ermos inside the practice.

The specialist letter to the referring GP

Written by the specialist after the first consultation and sent to the referring GP, with a copy to the patient. It tells the GP what the specialist found and what the GP is asked to do.

Harbourside Cardiology
Suite 4, 60 Alfred Street, Milsons Point NSW 2061 · (02) 5550 0178 · 3 November 2026

To: Dr Helena Voss, Kestrel Lane Medical Practice, 14 Kestrel Lane, Wollstonecraft NSW 2065

Re:

Dear Helena,

Thank you for asking me to see Maggie. I have summarised my findings below.

Findings.

Assessment.

Plan.

Medicines.

Social.

Yours sincerely,

An invented letter. The specialist and the clinic are fictional, and the Medicare and provider numbers are placeholders. All clinical values and doses are invented. Hover or tap any highlighted item for the analysis; the full list is below.

Specialist letter to GP drafter

An Agent+ that drafts the specialist's letter to the referring GP from consultation notes, results and the medication list. It uses the rooms' own letter template, puts findings and requested actions in separate parts, and flags dose changes and plan details for the specialist to confirm.

What it does

  1. Reads the consultation notes, results and medication list
  2. Drafts findings and requested GP actions in the template
  3. Flags dose changes and plan details to confirm

What stays with your people. The specialist checks, edits and signs every letter, and decides what is communicated. The agent drafts; it does not send.

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
    Name, date of birth, address and Medicare number

    Four identifiers that tie the letter to one person. The specialist’s rooms hold them as a health service provider, so the $3 million turnover threshold does not take the practice outside the Privacy Act (section 6D(4)(b)). Privacy Act guide.

  2. 2
    The test results

    An imaging finding, an ECG change and an ejection fraction against a named patient. Health information is sensitive information under section 6(1) of the Privacy Act. Patient data guide.

  3. 3
    The assessment and family history

    A diagnosis, a smoking history and a father’s heart attack. The family detail is health information about someone else that also bears on her own risk. A summary an AI tool produces can overstate or drop a qualifier, and the diagnosis is the specialist’s own judgement. Patient data guide.

  4. 4
    The plan and the travel advice

    The procedure, its date and place, and advice tied to a flight on 20 November. It says when she will be in hospital and when she may be away from home. Patient data guide.

  5. 5
    The medicine changes

    Doses, a new drug and an allergy. A wrong dose in a drafted letter becomes a wrong dose in a patient’s hands, so the specialist and the GP are the ones who check it. Ahpra says the practitioner remains responsible for safe, quality care and must apply human judgement to AI output. Patient data guide.

  6. 6
    A bereavement and a named daughter

    A fear, a late husband and a daughter named as support. Good medical practice 4.4 asks doctors to treat information about patients as confidential and to share it appropriately for their healthcare. Patient data guide.

  7. 7
    The requests to the GP

    Tests to repeat and when. Clinical instructions that make sense only with the rest of the letter. Patient data guide.

  8. 8
    The distribution instruction

    Who gets a copy and who does not. The patient has asked that her daughter is not sent letters, and the practice has to be able to follow that. HPP 11 in the NSW Health Records and Information Privacy Act and APP 6.1 limit disclosure for another purpose without consent or an exception. Privacy Act guide.

  9. 9
    The signatory and provider number

    A register name and a billing identifier, tied here to this patient.

  10. 10
    The closing line

    A routine courtesy that names no one. Replying to the referring doctor about the patient’s care is generally the purpose the information was collected for. Sending the letter to an AI tool is a different recipient, and APP 6.1, APP 6.2(a)(i) and HPP 10 and 11 have to be answered separately for it.

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: Referral letters on Ermos.

Common questions

Do ChatGPT and Claude use the information in a specialist letter to a GP?
A pasted letter 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 letter carries Maggie Ellison's Medicare number, a stress echocardiogram result, a plan for coronary angiography on 12 November and a note that she has been frightened of hospital since Harold died in 2022.
Can we use ChatGPT or Claude for a specialist letter to a GP if we remove the patient's name?
Not safely, because the letter still describes one woman. 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. A woman in Waverton with a stress test result, an angiogram on 12 November and a daughter called Claire is easy for people who know her to pick out. Her results and diagnosis are health information, which is sensitive information under section 6(1) of the Privacy Act. 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.
Can ChatGPT or Claude remember the details in a specialist letter to a GP later?
It depends on the account and its settings. Both products also have memory features that can carry details from one conversation into later ones, and on a personal account each user decides whether they are switched on (ChatGPT, Claude). This letter holds a penicillin allergy, new doses of atorvastatin and metoprolol, and her request that letters are not sent to her daughter. Those details could resurface in a later conversation about something else.
How can a specialist clinic use AI for a specialist letter to a GP 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 letter is drafted from the clinical record and the rooms' own letter template. The cardiologist checks every result, dose and date against the record and signs it.
Does a specialist need the patient's consent to write back to the referring GP?
APP 6.1 limits using or disclosing information for a purpose other than the one it was collected for. A reply to the referring doctor about the patient's care is usually the purpose it was collected for, so that is not the problem. Pasting the letter into an AI tool is another recipient and needs its own answer under APP 6.1 and 6.2.
Does the NSW health records law apply to a private specialist?
Yes. The Health Records and Information Privacy Act 2002 applies its health privacy principles to organisations that hold health information, including private ones. HPP 10 limits use and HPP 11 limits disclosure for a secondary purpose unless the person consents or the purpose is directly related and the person would reasonably expect it, or another listed exception applies.
What does Ahpra say about a specialist using AI to draft a letter?
Ahpra says the practitioner remains responsible for delivering safe and quality care and must apply human judgment to any AI output. It also asks practitioners to understand where data is located and stored and whether patient data is used to train the model, because those facts bear on confidentiality.

See Dominion answering from your own documents.

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

Book a demo