Claims chronologies and AI data leakage: what a claim file gives away in ChatGPT or Claude.

Claims handling became a financial service under the Corporations Act on 1 January 2021, with licensing applying in full from 1 January 2022, and a claim file is the most sensitive thing a brokerage holds: an injured person, what happened to them, what the witnesses said, what the doctors found, and what the insurer thinks it will cost. Chronologies are also exactly the task staff give ChatGPT or Claude, because assembling one from a folder of emails is slow. This page marks every item in a liability claim chronology that the Privacy Act and the Code of Practice bear on, and shows how the same chronology is built on Ermos inside the brokerage.

The claims chronology

Assembled by the claims team from the client’s notifications, the insurer’s correspondence and the evidence, to manage the claim and brief the client.

Claim chronology
Insured: · Insurer: Meridian General, claim · Prepared by · 22 October 2026

  1. 14 March 2026.
  2. 14 March.
  3. 17 March. Claim notified to Meridian under public liability.
  4. 2 April.
  5. 9 June.
  6. 23 July.
  7. 22 October.

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

What leaves the brokerage

  1. 1
    The insured and the claim

    A named client and the insurer’s claim number. A live claim, identified. Privacy Act guide.

  2. 2
    The injured person and the injury

    A named man, his age, his employer and a serious injury. Health information about a third party, which is sensitive information under the Privacy Act, and the most personal item in the file. Patient data guide.

  3. 3
    Witness statements, including an admission

    What two named people said, including a statement that a safety guard was removed. Evidence in a liability claim and a regulator’s investigation, attributed by name. Privacy Act guide.

  4. 4
    The reserve

    The insurer’s estimate of what the claim will cost. Confidential to the insurer and the client, and decisive in any negotiation.

  5. 5
    The medical report

    A surgeon’s findings, a permanent impairment, and a psychological referral, about someone who is not the client. The strictest category of information the Privacy Act recognises. Patient data guide.

  6. 6
    The letter of demand

    The claimant’s lawyers, the amount claimed and the allegation. The other side’s position in a dispute. Privacy Act guide.

  7. 7
    The indemnity problem and the lapsed policy

    The insurer’s possible exclusion argument and a lapse in the client’s workers compensation cover. The two facts most damaging to the client, and to the brokerage if the lapse was its oversight.

  8. 8
    The broker’s assessment

    That indemnity is at risk, that the director may be personally exposed, and that the client does not yet know. The brokerage’s judgement, and a note that it has not yet been shared.

  9. 9
    The claims handler

    Ties the file to a named staff member, in a service that has been regulated as a financial service since 2021, with a licence required from 1 January 2022.

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 brokerage 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 insurance brokerages. See how this works in practice: Claims chronologies on Ermos.

Common questions

Do ChatGPT and Claude use the information in a claims chronology?
Both receive all of it on the AI company's servers once it is pasted. 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. A claims chronology hands over an injured person's name, a surgeon's report and the insurer's reserve at once.
Can we use ChatGPT or Claude for a claims chronology if we remove the injured person's name?
No, because the rest of a claims chronology still points to one person. 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 34-year-old labour hire subcontractor with a partly amputated finger and a named hospital is recognisable to anyone who knows the case. 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.
Does a ChatGPT Business or Claude Team plan make it safe to use for a claims chronology?
Not on its own. 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. In a claims chronology that means the medical findings, the witness statements and the letter of demand still reach the AI company, whatever the plan. Our comparisons with ChatGPT Enterprise and Claude weigh the trade-offs.
How can a brokerage use AI for a claims chronology 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 chronology is built from the claim folder, with each event cited to the email or report it came from. The claims handler checks it against those documents and signs it off.
The injured person is not our client. Do our obligations cover him?
The Privacy Act covers personal information a brokerage holds, whoever it is about, once the brokerage is bound by the Act. Any brokerage with annual turnover above $3 million is, and some smaller ones are too. His medical report is sensitive information under it. The brokerage holds it only because it handles the client's claim, which is exactly the kind of limited purpose the Act expects to be respected.
Does claims handling being a financial service change anything?
Claims handling and settling is a financial service under the Corporations Act, and since 1 January 2022 the licensing requirement has applied in full. The obligation to provide financial services efficiently, honestly and fairly applies to it, and ASIC's guidance (INFO 253) describes that as handling claims in a timely way, fairly and transparently, and in a way that supports consumers. The guidance does not mention drafting tools, but a chronology drafted on a frontier model provider's servers, with the insurer's reserve and the client's indemnity problem in it, sits uneasily with looking after the claim for the client.
How is a chronology built on Ermos?
The claim folder is indexed on the unit in your office and the chronology is built from it, each event cited to the email or report it came from. The claims handler checks and completes it. Nothing about the claimant, the witnesses or the insurer's position leaves the brokerage.

See Dominion answering from your own documents.

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

Book a demo