Lender letters and AI confidentiality: what a capacity letter gives away in ChatGPT or Claude.

Liaising with a client’s bank is ordinary work for an accounting firm and unusually exposed work for AI confidentiality. A letter to a lender states the client’s income and structure in the firm’s own name, which makes it both a disclosure of the client’s affairs and a representation the firm is accountable for. This page marks every item in a typical letter supporting a loan application, and shows how the same letter is drafted on Ermos inside the practice.

The accountant’s letter to a lender

Requested by a lender or broker when a self-employed client applies for finance and the lender wants an accountant’s confirmation of income.

Northgate Partners
Chartered Accountants


Attn:

Re: ,

We have acted as accountants and tax agents for since 2019. We provide the following information at the client’s request and with his written authority.


An invented letter. The ABN is the ATO’s published example value. Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the firm

  1. 1
    The bank and the banker

    Which lender the client is dealing with and who at the bank. Information about the client’s financial relationships, and a named third party. APES 110 confidentiality guide.

  2. 2
    The client and the application

    Name, entity, ABN and the lender’s application reference. Enough to identify a live credit application. APES 110 confidentiality guide.

  3. 3
    Two years of taxable income

    Figures from lodged returns, stated in the firm’s name. Confidential financial information under APES 110 R114.2. APES 110 confidentiality guide.

  4. 4
    Personal income, including the spouse’s

    Two individuals’ income. The spouse’s is her personal information, disclosed to support her husband’s application. Privacy Act guide.

  5. 5
    Every existing debt

    Lenders, limits, balances and a Division 7A loan under repair. A complete liabilities picture, which is also a complete fraud target. APES 110 confidentiality guide.

  6. 6
    The trading downturn and the forecast

    A lost customer, a revenue fall and management’s recovery story. Market-sensitive to the client and material to the lender. TPB Code guide.

  7. 7
    The firm’s disclaimer

    The paragraph the firm relies on to limit its exposure. Professional bodies caution members about capacity letters; the letter’s wording is the firm’s protection, and it is now on a third party’s servers with the figures it qualifies. TPB Code guide.

  8. 8
    The partner

    The representation is made in a named practitioner’s name, which is what gives it weight to the lender and exposure to the firm. TPB Code guide.

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 firm 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 accounting firms. See how this works in practice: Client letters on Ermos.

Common questions

Do ChatGPT and Claude use the information in an accountant's letter to a lender?
It depends on the plan the user is on. 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. The letter is also sent to the AI company to be processed. This letter carries two years of taxable income, every debt and the partner's representation.
Can we use ChatGPT or Claude for an accountant's letter to a lender if we remove the client's name?
No, since a letter without the client's name still identifies one applicant. 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. An application number, a named lender and banker, a Division 7A loan and a customer that entered administration in July describe one business. The spouse's separate income is personal information in its own right. 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.
Can ChatGPT or Claude be trusted to draft an accountant's letter to a lender?
Not without a qualified person checking it against the lodged returns and the loan schedules. A language model can produce fluent text that changes a figure or a date. It can also add a detail the source never contained, or cite a reference that looks real but is not. A lender relies on this letter in the partner's name. The taxable income for each year, the facility balances and the Division 7A loan do the most damage if wrong. The forecast and the disclaimer wording are the hardest to repair if altered. The partner should check each of them before putting a name to the letter.
How can an accounting practice use AI for an accountant's letter to a lender 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. Ermos drafts the letter from the client's lodged returns and loan schedules, with the practice's standard wording as the template. The partner checks each figure and signs it before it goes to the bank.
The client authorised us to tell the bank. Does that cover AI?
The authority covers disclosure to the lender. It does not extend to disclosure to a frontier model provider used to draft the letter, which is a separate third party. Under APES 110 AUST R114.3 the client's authorisation has to cover the actual disclosure, and under the TPB Code the client's permission is specific to the recipient.
Why does the firm's own exposure matter here?
A lender letter is a representation in the firm's name. The disclaimer wording is how the firm limits its liability, and the figures are what the lender relies on. Both sitting on a third party's servers, outside the firm's control, is a risk to the firm as much as to the client.
How is this drafted on Ermos?
From the lodged returns and the loan schedules on your own network, with the firm's standard wording as the template. The letter is the same. The difference is that the client's income and debts never leave 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 accounting firms AI for accountants APES 110 and AI All professions