Cyber insurance proposals and AI leakage: what a cyber questionnaire gives away in ChatGPT or Claude.

A cyber questionnaire asks a client to describe, in its own words, every weakness in its systems. Brokers help complete it, and pasting the answers into ChatGPT or Claude is a tempting way to tidy the wording or check them against the insurer’s questions. The completed form says how many patient records the client holds, where they are backed up, who can log in remotely, how supplier payments are checked and what has already gone wrong. This page marks every item in a realistic dental practice proposal that the Insurance Contracts Act and the Privacy Act bear on, and shows how the same form is completed on Ermos inside the brokerage.

The cyber insurance questionnaire

Completed with the client before an insurer will quote cyber cover, and signed by a director with a declaration that the answers are true and complete.

Cyber insurance proposal
Anchor Underwriting · via Pacific Harbour Insurance Brokers · 30 October 2026

Applicant
Business and turnover
Personal information held
Systems and data location
Multi-factor authentication
Backups
Payment controls
Incidents in the past three years, insured or not
Declaration
Broker note

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

What leaves the brokerage

  1. 1
    The applicant, the address and the signatory

    The company, its trading name, the premises and the director who signed. Personal information about the director, and the identity of the business behind every answer below. Privacy Act guide.

  2. 2
    Size and turnover

    Staff numbers and annual turnover. Confidential financial information, and the turnover figure also bears on whether the Privacy Act small business exemption applies to a business (a private sector health service provider is covered whatever its turnover). Privacy Act guide.

  3. 3
    What personal information is held

    A count of patient records and the categories in them, including Medicare numbers and medical histories. It tells an attacker the value of the target and tells the insurer the scale of a possible breach. Health information is sensitive information under the Privacy Act. Patient data guide.

  4. 4
    Where the data lives

    Which software, which server and an offshore cloud backup. A map of where to attack, and a disclosure question for the practice about patient data held overseas. Privacy Act guide.

  5. 5
    The multi-factor authentication answers

    Which accounts have MFA and which do not, including a shared password for remote access. Published cyber proposal forms ask about MFA by service (email, remote access and privileged accounts). The answers are the practice’s weakest points in writing.

  6. 6
    The backup weaknesses

    A restore test from 2023 and a cloud copy that syncs and can be overwritten. Forms commonly probe whether a backup is truly separate from the live system. These are the facts an attacker would most like to know.

  7. 7
    The payment controls

    No call-back rule on supplier bank detail changes. Proposal forms commonly ask about a secondary check on changes of banking details, and the loss in the next item is what its absence looks like.

  8. 8
    Past incidents and the open loss

    Two losses, one of them days old and uninsured, with a police report and a bank investigation. Proposal forms commonly ask about incidents in the past three years whether insured or not. Disclosure goes to the insurer under section 21 of the Insurance Contracts Act, which is not the same as telling an AI company.

  9. 9
    The declaration wording

    The insurer’s standard declaration. It reflects the duty of disclosure in section 21 and the remedies in sections 28 and 60 of the Insurance Contracts Act, and it is the same text on every form of this kind, with no client data in it.

  10. 10
    The broker’s note on the MFA attestation

    The broker’s advice not to sign an attestation that is not yet true, and that a director has not seen the answers. Advice the Code of Practice describes as honest and in the client’s best interests (clauses 3.1 and 5.3).

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.

Common questions

Do ChatGPT and Claude use the information in a cyber insurance questionnaire?
Anything pasted or uploaded 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. A completed questionnaire lists patient Medicare numbers and medical histories, a shared password for remote access and a $61,000 loss from 19 October.
Is it safe to paste a cyber insurance questionnaire into ChatGPT or Claude?
No, because the answers set out the practice's weakest points in writing. 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. The missing MFA on remote access, a restore test last done in 2023, a cloud copy that can be overwritten and supplier bank changes confirmed by email only read as a map of where to attack. The director's signature ties them to one practice.
Can ChatGPT or Claude remember the details in a cyber insurance questionnaire later?
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 this form, that means the record count of about 9,400 patients, the shared remote access password and the 19 October loss could be carried into later conversations where they have no business being. The brokerage cannot control a setting on an account that belongs to a staff member.
How can a brokerage use AI for a cyber insurance questionnaire 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 answers are drafted from the client file and the brokerage's own records. The account manager checks each answer with the client, and a director signs the declaration only when every answer is true.
What do cyber insurance proposal forms usually ask?
Published Australian forms ask about multi-factor authentication for email, remote access and administrator accounts, whether a second check is made on payment requests and changes of banking details, how many personal records are held, and any incident in the past three years whether insured or not. The answers become the basis on which the insurer decides whether to offer cover.
What happens if the answers are wrong?
The Insurance Contracts Act 1984 puts a duty of disclosure on the insured before a contract is entered into (section 21) and sets out what the insurer can do where it is breached: avoid the contract if the failure was fraudulent, or otherwise reduce its liability for a claim (section 28), and cancel the contract (section 60). That can affect a later claim, so the broker's job is to make sure the answers are accurate and complete before a director signs.
Does the Privacy Act cover a dental practice with turnover under $3 million?
Yes. The OAIC lists private sector health service providers among the small business operators the Act covers regardless of turnover. A dental practice holding patient records is therefore bound by the Australian Privacy Principles, including the one on keeping personal information secure.

See Dominion answering from your own documents.

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

Book a demo