Healthcare Practice Staff Group Chats: Use case and AI Confidentiality

Most clinics have a team group chat on staff phones. It starts as a roster tool and turns into the fastest way to ask a colleague a question, which means patient names, reasons for rebooking, a rude phone call and a result that has not arrived end up in it. When a manager wants a summary of the week or help with a roster dispute, the thread is pasted into ChatGPT or Claude. This page marks every item in a routine staff group chat that the Privacy Act bears on, with the RACGP and Ahpra points on messaging, and shows how the same thread is handled on Ermos inside the practice.

The practice staff group chat

Kept by the whole team on personal phones, mostly for rosters, swaps and quick questions. The practice may not have a policy on what can be posted, and the messages are usually outside the clinical record.

Group chat:
Messages on staff phones ·

Tue 3 Nov, 8.04am Bec:

Tue 3 Nov, 8.11am Meera: Fri 9.30 is fine.

Tue 3 Nov, 11.15am Bec:

Tue 3 Nov, 11.19am Sam:

Wed 4 Nov, 8.55am Sam:

Wed 4 Nov, 9.20am Callum: Mum brought the form in. Can someone type it into the record? Nothing goes to dad.

Thu 5 Nov, 4.30pm Helen:

Fri 6 Nov, 9.05am Sam:

Fri 6 Nov, 9.12am Helen:

An invented group chat. All names are fictional and no app is named. Hover or tap any highlighted item for the analysis; the full list is below.

Staff thread roster summariser

An Agent+ that reads a team thread the practice chooses to give it and produces a tidy list of roster swaps and open questions. Using the practice's own communication policy, it flags messages that name a patient or hold clinical information so a person can move them to the proper place.

What it does

  1. Reads the thread and the practice's communication policy
  2. Lists roster changes, swaps and open questions
  3. Flags messages that mention patients or clinical details

What stays with your people. The practice manager reviews the list and decides what is recorded or removed. The agent summarises; it does not delete or post.

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 group and its members

    Names seven staff on personal phones. The Privacy Act treats an act done by a person employed by an organisation in the performance of their duties as the organisation’s act (s 8(1)(a)), so what staff post about patients in the course of their work is treated as the practice’s own act, whichever phone it is posted from. The Act reaches a private practice of any size that provides a health service and holds health information (s 6D(4)(b)). APP 11.1 requires reasonable steps to protect personal information from misuse, interference, loss and unauthorised access or disclosure, and APP 11.3 says those steps include technical and organisational measures. Privacy Act guide.

  2. 2
    A patient moved and why

    A patient’s initial, the clinician he sees, the time and his reason for rebooking, which, to anyone who knows Meera is the practice psychologist, shows that he sees one and that he does not want his manager to notice. Information about a health service provided or to be provided is health information (Privacy Act s 6FA(a)(iii)). Patient data guide.

  3. 3
    A caller who abused reception, and his warfarin status

    A patient named by surname, that he rang about his wife’s results, that he is on a warfarin clinic list and that a doctor has spoken to him. It reveals his treatment and records a complaint about conduct. Information that links a named patient to a health service is health information. Patient data guide.

  4. 4
    A result to chase and an instruction not to call

    A pregnant patient, a pathology laboratory and a test, with an instruction not to ring her until a doctor has seen the result. Health information is sensitive information (s 6(1)). The RACGP advises that clinical information be sent from within the clinical software using secure electronic messaging. Patient data guide.

  5. 5
    A photo of a child’s form, named for the child

    A file name that carries a child’s name and the service, a form behind it and a note that a parent must not be told. Ahpra’s social media guidance asks practitioners to take care when sharing comments or photos and says a personal account does not remove their obligations. Patient data guide.

  6. 6
    A caller about a clinic incident

    Names a patient, her daughter and a fall that is still under review, and records what reception said. Sensitive to the patient and to the practice, and in a place a review will not look for it.

  7. 7
    The practice’s own reminder

    A rule that results and clinical matters go through the clinical software and stay out of the chat. Generic wording that identifies no patient.

  8. 8
    A new member added

    Names a covering clinician and her start date. Adding a person changes who can read what is posted afterwards. Whether earlier messages are visible depends on the app, so the practice has to check how its own tool behaves.

  9. 9
    Disappearing messages

    A setting that deletes messages after seven days. It means the practice cannot show later what was said about a patient. Whether a particular message forms part of a patient’s health record is a question for the practice’s records policy and is not decided here.

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.

Common questions

Do ChatGPT and Claude use the information in a staff group chat?
A pasted thread 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 chat names a patient who moved a psychology booking so his manager would not notice, a man on the warfarin clinic list who abused reception, and a photo of a child's school form with a note that his father is not to be told.
Can we use ChatGPT or Claude for a staff group chat if we remove the names?
Not safely, because an initial, a clinician and a time still link a person to a health service. 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 patient moving a Thursday 6pm slot with Meera, or a pregnant patient waiting on a glucose test result, is recognisable to colleagues and to anyone who knows them. Information linking a named patient to a health service is health information, which is sensitive information. 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.
What if staff use their own ChatGPT or Claude account to summarise a staff group chat?
The practice cannot see or manage what a manager sends from a personal account. 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. Both companies can also review content that is flagged for safety or policy reasons. Section 8(1)(a) of the Privacy Act treats an act of an employee in performing their duties as the organisation's act, so a thread pasted to prepare a weekly summary is still the practice's concern.
How can a medical practice use AI to summarise a staff group chat 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 summary is drafted from the practice's own roster and records, with clinical matters handled in the clinical software as Helen's reminder in the chat says. The practice manager checks it before it is shared.
Can clinic staff use consumer messaging apps for patient information?
The Privacy Act has no rule about particular apps. APP 11.1 requires reasonable steps to protect personal information from misuse, loss and unauthorised access or disclosure, and APP 11.3 includes technical and organisational measures. The RACGP advises that clinical information go through the practice's clinical software using secure electronic messaging, and that a practice's policy cover patient consent for electronic transmission.
Is the practice responsible for what staff post on their own phones?
Under section 8(1)(a) of the Privacy Act, an act done by a person employed by an organisation in the performance of their duties is treated as the organisation's act. A message about a patient posted for work is therefore the practice's concern. Ahpra's social media guidance adds that a personal account does not remove a practitioner's professional obligations.
Does a patient's name in passing count as health information?
It can. Section 6FA of the Privacy Act counts information about a health service provided or to be provided to an individual, and other personal information collected in providing a health service, as health information. A rebooking that names a patient and a clinician, or a note that a result is awaited, links a person to a health service. Health information is sensitive information under section 6(1).

See Dominion answering from your own documents.

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

Book a demo