Client call transcripts and AI confidentiality: what a recorded call with a client gives away in ChatGPT or Claude.

Calls with clients are now recorded and transcribed by AI note-takers, and the transcript is pasted into ChatGPT or Claude to produce a file note, a summary for the partner or a draft statement. A transcript is the opposite of a polished letter. It holds the client’s anxieties, the lowest figure she will accept, the solicitor’s private opinion of the other side and the things the client admits she should have said earlier. This page marks every line of a short transcript that rule 9 and privilege bear on, and shows how the same call is summarised on Ermos inside the firm.

The client call transcript

Produced automatically after a recorded call, then summarised into a file note or used to prepare the next conference.

Call transcript
· Mon 26 Oct 2026, 4.18pm · 46 min (excerpt)

[00:00:06] Hamish Reid: Marguerite, it’s Hamish. I’m recording this call for the file. Are you happy with that?

[00:00:11] : Yes, that’s fine.

[00:01:40] Marguerite Soto:

[00:04:25] Marguerite Soto:

[00:05:02] Hamish Reid:

[00:12:10] Marguerite Soto:

[00:17:50] Hamish Reid:

[00:23:05] Marguerite Soto:

[00:23:30] Hamish Reid: That is below what we discussed. I’d want you and Eleanor in a conference before anything under $126,000.

[00:31:40] Marguerite Soto:

[00:32:05] Hamish Reid: Don’t delete or change anything else, and don’t try to restore it yourself. I’ll arrange that with you properly.

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

What leaves the firm

  1. 1
    The client’s voice and name

    A name against a recorded voice and a matter. Personal information, and a client confidence from the first line. Rule 9 confidentiality guide.

  2. 2
    Health, medication and what she keeps from others

    A medication change, a symptom and a secret kept from the children’s father. Health information is sensitive information under the Privacy Act, which applies to a law practice only if the practice is covered. Privacy Act guide.

  3. 3
    A job interview

    A named prospective employer and a start date. It bears on the compensation claim, and it is a fact the client has not told the other side. Legal privilege guide.

  4. 4
    The solicitor’s advice on mitigation

    What a new job does to the claim. Legal advice given in confidence for the dominant purpose of the matter, which is what privilege attaches to. Legal privilege guide.

  5. 5
    A relative’s view and the figures

    What the client’s sister thinks she should accept, with two numbers. Third-party opinion and the client’s settlement thinking. Legal privilege guide.

  6. 6
    The firm’s candid view of the other side

    A private opinion that opposing counsel is bluffing. The kind of remark made on the phone that never reaches a letter. Legal privilege guide.

  7. 7
    The settlement floor

    The lowest figure the client will accept and the non-money terms she wants. If the other side saw this line, the negotiation would be over. Legal privilege guide.

  8. 8
    A deleted text and a possible preservation problem

    An admission about a document that matters to the case and where it might be recovered. The most sensitive line in the transcript. Legal privilege guide.

  9. 9
    How the call was recorded

    The recording system and the transcription tool, named in the transcript header. The firm’s tool choices, and the reminder that a transcript is itself a document. 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

Ermos Dominion

The document is drafted on a unit in your office, from your own templates and the matter 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 law firms.

Common questions

Do ChatGPT and Claude learn from a client call transcript?
Yes, a pasted transcript 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. This excerpt holds a medication change, a $115,000 figure that is the lowest the client will accept and a deleted text message she mentions late in the call.
Can we use ChatGPT or Claude for a client call transcript if we remove the names?
No, because a transcript identifies the client through what she says. 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. A finance manager interview at a named company starting in January, a dose change from her doctor and a deleted text from 2024 describe one person. 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. Health information is also sensitive information under the Privacy Act, which applies to a law practice only if the practice is covered.
Is it safe to use an AI note-taker or meeting recorder for a client call transcript?
Not until the firm has answers on where the recording is stored, whether the tool's vendor trains a model on it and who can review it. Recording and transcription tools send the audio or transcript to the tool's vendor, so those questions apply to every call. 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 transcript's header shows the call was recorded on the firm's phone system and transcribed by an AI note-taker.
How can a law firm use AI for a client call transcript 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 file note and summary are drafted from the transcript and the firm's own records, and the solicitor checks the summary against the transcript and signs the note.
Is a recorded call with a client privileged?
Privilege attaches to confidential communications made for the dominant purpose of giving or receiving legal advice, and a transcript is a record of those words. A recording or transcript of such a call can therefore be privileged in the same way a file note can be, to the extent the words meet that test. The question for a firm is whether sending it to a third party's tool is consistent with keeping it confidential.
Why is a transcript riskier than the file note that follows it?
A file note records what the solicitor chose to write. A transcript records everything, including the client's asides, her lowest figure and the solicitor's candid remarks about the other side. Those are the lines that never reach a letter, and they are the ones the other side would most like to read.
Can a call be summarised without the transcript leaving the firm?
On Ermos, yes. The recording and the transcript sit on your own network, and the unit produces a file note, a list of actions and a summary for the partner, each citing the lines it drew on. The solicitor reads the transcript against the summary before the note goes on the file.

See Dominion answering from your own documents.

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

Book a demo