Underwriting emails and AI data leakage: what liaising with an insurer gives away in ChatGPT or Claude.

Liaising with an insurer about a client’s application is the most medically detailed writing an adviser does. The underwriter needs the diagnosis, the treatment, the specialist and the dates, and the adviser needs to present them well, which is why the email gets drafted in ChatGPT or Claude. This page marks every item in a routine underwriting email that the Privacy Act protects, with the health information treated as sensitive information, and shows how the same email is drafted on Ermos inside the practice.

The email to the insurer

Sent to an insurer’s underwriter when an application is referred for further information, with the client’s authority.

To: Underwriting, Meridian Life
From:
Subject: Application : further information

Further to your request of 20 October, with the client’s authority:

  1. Thyroid condition.
  2. Mental health.
  3. Occupation and income.
  4. Cover sought.
  5. Family history.

An invented email. Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the practice

  1. 1
    The client and the application

    Named, with the insurer’s reference. A live underwriting file. Privacy Act guide.

  2. 2
    Diagnosis, medication, doctors and results

    Health information in full, with two named practitioners and test results. Sensitive information under the Privacy Act, with the strictest handling rules. Patient data guide.

  3. 3
    Mental health history

    Six sessions with a psychologist and the context the client asked to be considered. The most private paragraph in the email, disclosed to an insurer with care and to a frontier model with none. Patient data guide.

  4. 4
    Occupation and income

    Employment and financial information, with the tax return attached. Privacy Act guide.

  5. 5
    The cover sought

    What the client is insuring for and how much. Financial information.

  6. 6
    Family medical history

    Health information about two other people, one of them deceased. The mother’s is sensitive information about someone who is not the client. The father’s sits outside the Privacy Act in its own right, though it can still reveal the client’s own health risks. Patient data guide.

  7. 7
    The client’s preference and a deadline tied to her husband’s job

    A negotiating position, and a hint at the husband’s confidential job change. Privacy Act guide.

  8. 8
    The adviser

    Ties the email to a named adviser bound by the Code of Ethics.

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 client file. 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 financial advice practices. See how this works in practice: Advice drafting on Ermos.

Common questions

Do ChatGPT and Claude use the information in an underwriting email to an insurer?
Yes, the whole email is sent to the AI company's servers and processed there. 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. It holds a diagnosis with two named doctors, six sessions with a psychologist and the medical history of two relatives.
Is it safe to paste an underwriting email into ChatGPT or Claude?
No, because an underwriting email is the most medically detailed writing an adviser does. 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. Removing the client's name still leaves the insurer's application reference, the diagnosis date and an appointment with a named specialist, which point to one person. The diagnosis and the psychologist's sessions are health information, which is sensitive information under the Privacy Act, and so is the mother's diabetes.
Does deleting the chat remove an underwriting email from ChatGPT or Claude?
Not completely. Deleting a chat removes it from the user's history, but OpenAI and Anthropic each say they can keep a deleted chat on their systems for up to 30 days, and longer where safety or legal obligations apply (OpenAI, Anthropic). 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). For an underwriting email, a diagnosis, a psychologist's sessions and two relatives' medical histories could sit beyond the practice's reach even after the adviser has deleted the chat.
How can an advice practice use AI for an underwriting 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 fact find, the medical documents and the application in the practice's usual format. The adviser checks it against them before it goes to the insurer.
The client authorised disclosure to the insurer. Does that cover AI?
The authority covers the insurer, for underwriting. A frontier model provider used to draft the email is a separate recipient the client never named. Under APP 6, disclosing sensitive information for a purpose other than the one it was collected for needs the client's consent or one of the exceptions in APP 6.2, such as a directly related purpose the client would reasonably expect. Under APP 8, an overseas provider adds the cross-border questions.
Is health information different from the rest of the file?
Yes. Health information is sensitive information under the Privacy Act, which means stricter rules on collection, use and disclosure than ordinary personal information. Underwriting emails are where most of an advice practice's health information is written down in prose.
How is an underwriting email drafted on Ermos?
From the fact find, the medical documents and the application on your own unit, in the practice's usual format. The adviser checks it and sends it to the insurer. The client's medical history never leaves the practice to produce it.

See Dominion answering from your own documents.

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

Book a demo
Keep reading
AI data leakage in financial advice practices AI for financial advisers The Privacy Act, the APPs and AI All professions