Letters of demand and AI security: what an unpaid-invoice demand gives away in ChatGPT or Claude.

A letter of demand looks like the safest thing a law firm drafts, because the debtor is about to read it anyway. That is why staff paste drafts into ChatGPT or Claude to firm up the tone. The draft is built from the client’s confidential file, though, and the notes around it say how badly the client needs the money and what it will take instead. This page marks every item in a demand for unpaid fabrication invoices that rule 9 and privilege bear on, explains what each one exposes, and shows how the same letter is drafted on Ermos inside the firm.

The letter of demand

Sent to a customer that has not paid, usually after reminders have failed and before a statement of claim is filed.

An invented letter, shown as the internal draft with its note still attached. The ABN is the ATO’s published example. Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the firm

1. 1
   The debtor, its director and its address

   A named company with an ABN, a named director and a business address, in a debt dispute. Confidential to the client under rule 9 until the letter is sent, and the director’s name is personal information about an identifiable individual. Rule 9 confidentiality guide.

2. 2
   The contract and what was supplied

   The purchase order number, the product and the quantity. It identifies a commercial relationship the client has not published. Rule 9 confidentiality guide.

3. 3
   The client

   That a named company has instructed solicitors to pursue a named customer. A fact the client may not want known before the letter lands, and information confidential to the client under rule 9. Rule 9 confidentiality guide.

4. 4
   The invoices and the amount

   Invoice numbers, dates and a total of $412,860.00. The client’s receivables, and the figure the whole dispute turns on. Rule 9 confidentiality guide.

5. 5
   A personal guarantee

   A named individual’s personal liability and the date he signed. The client’s strongest lever against a person rather than a company, drafted from the client’s documents, and personal information about him. Privacy Act guide.

6. 6
   The debtor’s own words and its financier

   What the debtor wrote on 24 September, and who is financing it. The firm has chosen which of the debtor’s statements to rely on, which is litigation strategy, and the financier is a third party. Legal privilege guide.

7. 7
   The demand, the rate and the deadline

   Amount, interest and a date. Known to the debtor once the letter is sent, and until then the client’s confidential position. Rule 9 confidentiality guide.

8. 8
   The standard warning paragraph

   Wording that any demand carries. It names no client, party or figure.

9. 9
   The internal note

   How much the client needs the money, the date its bank reviews the facility and the figure it will accept instead of suing. Privileged client instructions: if this reaches the debtor, the negotiation is over. Legal privilege guide.

10. 10
    The partner

    Ties the demand to a solicitor bound by rule 9 and by the Law Society’s guidance on AI. Law Society AI guidance.

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

Part of AI data leakage in law firms. See how this works in practice: Litigation work on Ermos.

Common questions