Client database exports and AI data leakage: what a client list gives away in ChatGPT or Claude.

Every advice practice can export its client list from the planning software, and every practice does, for a mail-out, a review schedule, a fee check or a clean-up before an audit. The export is the most concentrated document the practice owns, because it is every client in one grid: name, date of birth, part of a tax file number, the fee, the risk profile and a free-text notes column written by the adviser. Staff paste it into ChatGPT or Claude to sort it or find the gaps. This page marks the items in an invented export that the Privacy Act, the TFN Rule and the Code of Ethics bear on.

The client list export

Exported from the practice’s planning software when the team needs a view of the whole book, for review scheduling, fee reconciliation, mail-outs or an audit. Standard 8 of the Code of Ethics requires records of clients, including former clients, to be kept complete and accurate.

fx=SUM(D2:D10)⁹

	A	B	C	D	E	F	G
1	Client	DOB	TFN (last 3)	Ongoing fee ($ pa)	Risk profile	Adviser notes	Next review
2	Harrington, Thomas James¹	14/03/1979¹	…782²	4,400³	Growth⁴	Redundancy in December (confidential). Holds Tesla shares against advice.⁵	21/01/2027
3	Harrington, Lucy Anne¹	02/08/1981¹	to come²	joint	Balanced⁴	Thyroid recurrence. Income protection accepted with an exclusion. Estate referral not to be copied to Tom.⁵	21/01/2027
4	Ellis, Margaret Joy¹	11/09/1955¹	…640²	3,300³	Conservative⁴	Widowed 19 October. Son Peter wants the money sorted quickly. She is unsure and tired.⁵	16/11/2026
5	Ellis, Graham Roy⁶	03/02/1951¹	…219²	closed	Conservative	Deceased 19/10/2026. Life claim notified. Super fund notified.⁶	n/a
6	Desai, Anil Kumar¹	07/06/1974¹	…905²	2,750³	Balanced	Income protection lapsed 12 September. Knee surgery in August and blood pressure tablets. Must disclose on reinstatement.⁵	30/11/2026
7	Brennan, Liam¹	22/11/1988¹	…317²	2,200³	Growth	Separated from Sophie. Wants nothing sent to the shared address.⁵	10/02/2027
8	Fitzgerald, Joan Mary⁷	19/04/1940¹	…058²	1,650³	Conservative	Memory concerns. Daughter Karen holds an enduring power of attorney that is yet to be activated. No instructions by phone alone.⁷	05/03/2027
9	Chen, Wei Ling¹	30/09/1969¹	…772²	3,850³	Balanced	Selling her dental practice in 2027, price confidential. Tax debt being paid by instalments.⁵	25/11/2026
10	Mahoney, Declan¹	14/01/1983¹	…126²	2,200	Growth	Debt agreement ended 2023. Rebuilding. Anxious about retirement.⁵	18/02/2027
11	Total			20,350⁹		Hidden columns: mobile, email, spouse, planning software ID⁸

Active clientsClosedPivot⁸

An invented client list. The tax file number digits are placeholders apart from the ATO’s example ending 782. Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the practice

Names and dates of birth

Nine named people, each with a date of birth. Personal information under the Privacy Act where it applies to the practice, and two of the standard facts used to verify identity. Privacy Act guide.

Frontier Models (ChatGPT, Claude)	Nine identities transmitted, generally offshore on personal plans
Ermos Edge	Stays on your instance in Australia; leaves only inside a permitted excerpt, per question, to the named provider
Ermos Dominion	Stays on the unit in your office

Tax file number digits

Part of a TFN beside a name and date of birth. The Privacy (Tax File Number) Rule restricts how TFN information may be used and disclosed. Whether a fragment is TFN information is a question for the practice, and a full TFN in the same file plainly would be. Privacy Act guide.

Frontier Models (ChatGPT, Claude)	Tax file number fragments with names and dates of birth transmitted
Ermos Edge	Stays on your instance in Australia. If it sits in a permitted excerpt, it goes only to a named Australian provider that answers and keeps nothing
Ermos Dominion	Stays on the unit in your office

Fees by client

What each client pays the practice each year. Commercial information about the practice and personal information about the client, and the basis on which the client has consented to ongoing fees.

Frontier Models (ChatGPT, Claude)	Every client’s fee transmitted
Ermos Edge	Stays on your instance in Australia; leaves only inside a permitted excerpt, per question, to the named provider
Ermos Dominion	Stays on the unit in your office

Risk profiles

How each client says they handle loss. The kind of detail Standard 8 expects in a complete and accurate client record, and the kind a practice keeps confidential.

Frontier Models (ChatGPT, Claude)	Every client’s risk profile transmitted
Ermos Edge	Stays on your instance in Australia; leaves only inside a permitted excerpt, per question, to the named provider
Ermos Dominion	Stays on the unit in your office

Candid adviser notes

A redundancy, a health condition, a separation, a tax debt, a business sale and a private request. The free-text column holds what clients said in confidence. Health detail is sensitive information under the Privacy Act. Patient data guide.

Frontier Models (ChatGPT, Claude)	What clients said in confidence transmitted; on consumer tiers it can train future models
Ermos Edge	Stays on your instance in Australia; leaves only inside a permitted excerpt, per question, to the named provider
Ermos Dominion	Stays on the unit in your office

A deceased client

A former client, with the date of death and the claims in train. The Privacy Act does not apply to the personal information of a deceased individual, but Standard 8 requires records of former clients to stay complete and accurate, and the surviving family appears in the notes.

Frontier Models (ChatGPT, Claude)	Not covered by the Privacy Act for a deceased person, but it is still a client record and it is transmitted
Ermos Edge	Stays on your instance in Australia; the deceased client’s record is still the practice’s to keep
Ermos Dominion	Stays on the unit; the deceased client’s record is still the practice’s to keep

A vulnerable client and who holds her power of attorney

A client with memory concerns, her daughter’s name and an instruction about phone contact. Personal and health information about a client who is easy to target, with the safeguard the practice put in place. Patient data guide.

Frontier Models (ChatGPT, Claude)	A vulnerable client and her safeguards transmitted
Ermos Edge	Stays on your instance in Australia; leaves only inside a permitted excerpt, per question, to the named provider
Ermos Dominion	Stays on the unit in your office

Hidden columns and other tabs

Hidden columns holding contact details, other tabs and a pivot table. A spreadsheet file can carry hidden columns, other sheets, comments and properties that stay out of sight in the grid and travel with the file.

Frontier Models (ChatGPT, Claude)	Hidden columns and other tabs travel with the file
Ermos Edge	Stays on your instance in Australia; leaves only inside a permitted excerpt, per question, to the named provider
Ermos Dominion	Stays on the unit in your office

A formula and a total

A sum of ongoing fees across the book, built from a visible range. The total says how much the practice bills.

Frontier Models (ChatGPT, Claude)	A sum of fees; harmless alone, revealing beside the practice’s name
Ermos Edge	Stays on your instance in Australia; leaves only inside a permitted excerpt, per question, to the named provider
Ermos Dominion	Nothing leaves; the total still needs checking against the fee agreements

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 practice 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 Edge	Ermos Dominion
Where the document is processed	Provider servers, generally offshore on personal plans	Your instance in an Australian data centre; answers generated in Australia	The unit in your office
Who can read it	Provider staff and systems, under policy	Your staff, within the boundaries you set. The named Australian provider processes only permitted excerpts, under contract, and keeps nothing	Your staff only
Is it kept	On the provider’s terms	Kept on your instance; the provider is barred from retaining it	Stays on the unit
Can it train a model	On consumer tiers, yes	No, by contract	Only your own unit learns from it, on your hardware

Part of AI data leakage in financial advice practices.

Common questions

Do ChatGPT and Claude use the information in a client list export?

Yes, everything pasted or uploaded from the export 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. The export puts nine clients' dates of birth, tax file number digits, fees and candid adviser notes in one grid, including a confidential redundancy and a client with memory concerns.

Is it safe to paste a client list export into ChatGPT or Claude?

No, because the export holds every client in one grid. 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. With the names gone, the dates of birth, the tax file number digits and the free-text notes on a redundancy, a separation and a tax debt can still point to real people. 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.

Can I upload a client list export to ChatGPT or Claude instead of pasting it?

It is no safer, because an uploaded file leaves the business as a whole file. It can carry other sheets, hidden columns, comments, tracked changes and file properties as well as the part you meant to share. The OAIC's guidance also tells organisations to review carefully the terms and settings that will apply to their use of any AI product. This export has a Closed tab, a Pivot tab and hidden columns of contact details, and all of them would travel with the visible list.

How can an advice practice use AI for a client list export 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. Questions about the book are answered from the practice's own client records. The adviser checks each answer against the client files and fee agreements before the team acts on it.

Is a partial tax file number still TFN information?

The Privacy (Tax File Number) Rule restricts how TFN information may be used and disclosed. Whether a few digits kept beside a name and date of birth count is a question for the practice and its licensee. A full TFN in the same file would clearly be covered, so many practices hold only the digits they need for the task.

What does the Code of Ethics say about client records?

Standard 8 of the Code of Ethics requires an adviser to ensure that records of clients, including former clients, are kept in a form that is complete and accurate. That is why a client list carries free-text notes and why a deceased or closed client stays on file. Completeness inside the practice is the obligation; where the list is sent is a separate question.

If a client list is exposed, does it have to be reported?

Under the Notifiable Data Breaches scheme, an entity covered by the Privacy Act must assess a suspected eligible data breach, which involves unauthorised access or disclosure that is likely to result in serious harm. A list with dates of birth, tax file number digits and health notes is the kind of disclosure that assessment exists for. The practice's own circumstances decide the outcome.

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 financial advice practices AI for financial advisers The Privacy Act, the APPs and AI All professions