Health fund claim emails and dental AI security: what a rebate query gives away in ChatGPT or Claude.

Dental practices write to health funds when a rebate is unclear, a claim is rejected or a patient asks what they will get back. The practice manager is under time pressure, the fund’s rules are fiddly, and the email is a natural thing to give to ChatGPT or Claude to draft and make polite. It carries the member’s number, the treatment and its fee, other patients in the same batch, the practice’s bank details and often something a patient said in confidence about their cover. This page marks every item in a routine fund email that the Privacy Act and the Code of conduct bear on, and shows how the same email is drafted on Ermos inside the practice.

The health fund claim email

Sent by the practice manager to a health fund’s provider team when a rebate needs confirming before treatment or a claim has been rejected. The patient has authorised the practice to claim, but only for the claim.

To: Provider Services, Kestrel Health Fund
From:
Subject: Rebate confirmation and three rejected claims, provider

Dear Provider Services,

I am writing about three matters, all for treatment we are about to provide or have provided.

1. Pre-treatment query.

2. Rejected claims.

3. Payment.

Kind regards,

An invented email. The practice, fund and patients are fictional; the member numbers, provider number, phone number and bank details are placeholders (ABN 51 824 753 556 is the ATO’s published example). Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the practice

  1. 1
    A named patient, her member number and the treatment

    Identity, fund membership and a procedure with its date and fee. Personal information, and health information because it describes treatment given in providing a health service. Patient data guide.

  2. 2
    A patient’s intention to cancel her cover

    What a patient said about her own cover, repeated to the fund. Confidential to her and to the practice, and sent to an AI tool it reaches a third party. Privacy Act guide.

  3. 3
    Three other patients in the same batch

    Names, member numbers and treatments for patients who are not the subject of the email. Their health information, in a message about someone else. Patient data guide.

  4. 4
    A patient’s request about contact

    Resubmission instructions and a request not to phone a patient at home while a pathology result is pending. Sensitive information about an investigation of a patient’s health. Patient data guide.

  5. 5
    Bank details and ABN of the practice

    Where the money goes. The practice’s own information, and the kind of detail used in payment-redirection scams.

  6. 6
    The provider number

    The practice’s identifier with the fund. Used to authenticate claims and not personal to any patient.

  7. 7
    The request for confirmation

    Generic wording asking the fund to confirm in writing. The practice’s own template and nothing about any patient.

  8. 8
    The sender

    A named practice manager with a direct line. Ties the message to the practice and to a person who handles patient data.

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 AI data leakage in dental practices.

Common questions

Do ChatGPT and Claude use the information in a health fund claim email?
Yes, a pasted claim email 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 email lists four patients by name with fund member numbers, a patient's plan to cancel her cover, and the practice's BSB and account number.
Is it safe to paste a health fund claim email into ChatGPT or Claude if we remove the patient names?
No, because the member numbers, dates and treatments still identify people. 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 fund member number beside a procedure on a named date, and a request not to ring a patient at home while a pathology result is pending, each point to one person. Three other patients in the same email have their health information exposed too. 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 for a health fund claim email?
The practice cannot see what was sent. 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. An email pasted from a personal account sits outside the practice's own records and controls, so nobody can say which member numbers, rejected claims or bank details left the practice.
How can a dental practice use AI for a health fund claim email 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 email is drafted from the patient files and the practice's own claim query template. The practice manager checks every member number, claim code and amount and signs it.
The patient authorised us to claim. Does that cover AI?
The authority covers the fund, for the claim. An AI company used to draft the email is a separate recipient the patient never named. Under APP 6 a practice may use or disclose personal information for a purpose other than the one it was collected for only in the cases the Act allows, such as consent or, for health information, a directly related purpose the patient would reasonably expect.
Why do other patients in the email matter?
Because a batch email about several rejected claims carries their health information as well as the subject patient's. Each patient is entitled to have their own information handled for their own claim. A paste to an AI tool puts all of them in front of a recipient none of them was told about.
Are practice bank details a privacy issue?
Not under the Privacy Act, which is concerned with personal information about individuals, but they are a fraud risk: payment-redirection scams use real details from real emails. A practice should treat bank details as sensitive business information and keep them out of tools that retain what they receive.

See Dominion answering from your own documents.

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

Book a demo