AI data leakage in mortgage broking: what leaves when client files go into ChatGPT or Claude.

A mortgage broker holds the documents a household would least like to see leave: payslips, bank statements, identity documents, credit reports and the reasons one loan was recommended over another. The National Consumer Credit Protection Act requires a broker to act in the consumer’s best interests (sections 158LA and 158LE), to give the consumer’s interests priority where they conflict with the broker’s, and to make reasonable inquiries and take reasonable steps to verify the consumer’s financial situation (section 117). The Privacy Act treats everything in the file as personal information, subject to its small business exemption. This series takes ten documents a brokerage produces, marks every protected item, and shows where each one goes when the document is pasted into ChatGPT or Claude. Each page ends with the same document on Ermos.

The documents, one at a time

FormThe loan application form

Two identities, employers, every account and card, the children, the property and a family gift. A household on one form.

See what leaves → SpreadsheetThe serviceability spreadsheet

Net income, shaded overtime, living expenses, every card limit and a surplus of a few hundred dollars. How close a family is to its limit.

See what leaves → NotesThe best interests file note

The clients’ objectives, the lenders compared, the commission difference and the broker’s candid impressions. Why this loan, in writing.

See what leaves → EmailThe lender submission email

The applicants, the deal, income, a family gift, a late payment, an earlier decline and the broker’s candid aside. The pitch, before the lender sees it.

See what leaves → RecordThe supporting documents summary

Identity scans, payslips, bank statements, a credit report and what the family spends money on. The paperwork a household would least like read by a stranger.

See what leaves → EmailThe pre-approval email to clients

The lender, the amount, the conditions, the savings advice and a private plan mentioned on a call. Good news, with the clients’ whole position in it.

See what leaves → EmailThe declined application email

The decline reason, the credit enquiry, an admitted error and the plan B. The worst day of the application, in writing.

See what leaves → TranscriptThe broker and client call transcript

Income, a job worry, a late payment, a family gift, a private plan and the broker’s advice, in the clients’ own words.

See what leaves → SpreadsheetThe commission and clawback spreadsheet

Every client with a loan amount and lender, the commission rates, trail income and the date each clawback ends. The brokerage’s book of business.

See what leaves → LetterThe complaint response

The allegation, the chronology, a quote from the call recording, a partial admission and the remedy, in a letter AFCA may read next.

See what leaves →

How the leakage happens

The mechanism is the same for every document. The whole text 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. On consumer tiers it can be retained and used to train future models. The provider’s staff and systems can reach it under the provider’s own policies. Once sent, the brokerage no longer controls it. Our on-premise vs cloud comparison sets out the four ways to run AI and where each one puts your documents.

The obligations in play

Each marked item on these pages links to the guide that covers its obligation:

The same documents on Ermos

Ermos Dominion drafts and searches on a unit in your office, so the documents never leave it. Ermos Edge keeps your documents on your own instance in an Australian data centre and sends only permitted excerpts, per question, to a named Australian provider. Every page in this series ends with both, item by item.

Common questions

Can mortgage brokers use ChatGPT or Claude with client information?
Only with care, because the credit laws and the Privacy Act both stay in play. 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 best interests duty and the inquiry and verification steps in section 117 apply to credit assistance however it is prepared. The Privacy Act covers the file unless the brokerage is a small business with turnover of $3 million or less, and commission income counts toward that figure. If a tool's servers are outside Australia, APP 8.1 can apply.
Do ChatGPT and Claude train on what staff paste in?
It depends on the account each person is using. 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. For a mortgage brokerage the question matters because one paste can carry two identities, payslips, account endings and a credit report.
Do business plans such as ChatGPT Business or Claude Team solve the problem for mortgage brokers?
Only in part. 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. For a brokerage, the open question is whether the clients agreed to their payslips and credit history going to that AI company. Our comparisons with ChatGPT Enterprise and Claude weigh the trade-offs.
What is the safest way for a mortgage brokerage to use AI with client documents?
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. For a mortgage brokerage, staying in control means the household's payslips, statements and credit reports stay where the brokerage can answer for them.
What is AI data leakage in a mortgage brokerage?
Client information leaving the brokerage's control through an AI tool. When a payslip summary, a serviceability sheet or a file note is pasted into ChatGPT or Claude it is transmitted to the provider's servers, can be kept on the provider's terms, can be used to train future models on consumer tiers, and can be reached by the provider's staff. For a broker the information is a household's income, debts, credit history and spending.
Which broking documents carry the most risk?
The supporting documents summary and the application form, because they gather identity, income, accounts and credit history in one place; the serviceability spreadsheet, because it shows exactly how close a household is to its limit; and the call transcript, because clients say things in a conversation that never reach the form. The pages in this series take each in turn.
Do the credit laws say anything about AI?
Not by name. The best interests duty in section 158LA and 158LE, the inquiry and verification steps in section 117, and the general conduct obligations in section 47 apply to credit assistance however it is prepared. The Privacy Act applies to the personal information involved, and from 10 December 2026 APP 1.7 to 1.9 require an APP entity's privacy policy to describe decisions made, or substantially and directly supported, by a computer program using personal information, where the decision could reasonably be expected to significantly affect an individual. The MFAA's 2024 discussion paper on AI is industry guidance.

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 for mortgage brokers The Privacy Act, the APPs and AI All professions