Migration call transcripts and AI data leakage: what a recorded client call gives away in ChatGPT or Claude.

The first call with a prospective client is where the facts come out: how they arrived, what went wrong, what they did that they are worried about. Agents record the call or run it through an AI note-taker, then paste the transcript into ChatGPT or Claude to produce a file note or an advice email. The transcript holds every word as spoken, including an admission about breaching a visa condition and the agent’s blunt view of the client’s chances. This page marks every item in a typical intake call transcript that the Code of Conduct and the Privacy Act bear on, and shows how the same transcript is handled on Ermos inside the practice.

The first client call transcript

Recorded or transcribed on the day of the first consultation. Section 55 of the Code requires an agent to make a contemporaneous written record of oral communications with a client about immigration assistance, including oral instructions and oral advice. Section 43 allows advice at an initial consultation without a service agreement.

Call transcript
Generated by the practice’s AI note-taker · Wednesday 28 October 2026 · 22 minutes

[00:00:41] Renata: This is Renata Kowalczyk-Bell at Kingfisher Lane Migration, MARN 0000000. Can you tell me who you are?

[00:01:10] Farid:

[00:02:30] Renata: And your visa history?

[00:02:44] Farid:

[00:05:12] Renata: Is there anything in your history that worries you?

[00:05:20] Farid:

[00:09:03] Renata: Your family circumstances?

[00:09:15] Farid:

[00:14:40] Renata:

An invented transcript. The client, family, employer and hospital do not exist, and the MARN is a placeholder. Hover or tap any highlighted line for the analysis; the full list is below.

What leaves the practice

  1. 1
    The client and his family

    A named man, his wife and the names of his children. Personal information, and minors are among those named. Privacy Act guide.

  2. 2
    Age, birthplace and citizenship

    A birth year, place of birth, nationality and where the family lived. Personal information, and details that can point to national and ethnic background. Privacy Act guide.

  3. 3
    Visa history

    How and when he arrived and what visa he holds now. Immigration history the practice needs and an adverse party would value. Privacy Act guide.

  4. 4
    An admission about working beyond visa conditions

    He says he worked about 60 hours a fortnight in cash while on a student visa. A statement against his own interest, made in confidence to the agent, and exactly what would trouble a character or compliance assessment. Privacy Act guide.

  5. 5
    A prior refusal

    A refused visitor visa for his brother. Personal information about a third person, and a refusal history. Privacy Act guide.

  6. 6
    His wife’s health

    A diagnosis, a treatment and a hospital. Health information about a third person, and sensitive information where the Privacy Act applies. Patient data guide.

  7. 7
    Income and savings

    Earnings, savings and where the savings came from. Financial information, with the source of the money being unusually sensitive. Privacy Act guide.

  8. 8
    The agent’s advice and futility warning

    Candid advice about his chances and a statement that the agent must advise him if assistance would be futile. Section 19 of the Code bars assistance an agent reasonably believes futile unless the client has been advised and has given a written statement that he still wishes to proceed.

  9. 9
    The consultation fee and the recording notice

    A fee for the initial consultation, which section 43 requires to be reasonable, and the agent’s statement that the call is recorded. A routine disclosure with no personal information about the client.

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 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 migration practices.

Common questions

Do ChatGPT and Claude use the information in a client call transcript?
Yes, whatever is pasted from the 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. The transcript records an admission about working 60 hours a fortnight in cash, his wife's diabetes and the agent's view that he cannot reach 65 points.
Is it safe to paste a client call transcript into ChatGPT or Claude if we remove his name?
No, because a transcript records every word as spoken. 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 45-year-old born in Kabul, with three children and a wife treated at Blacktown Hospital, stays identifiable without a name, and the admission about cash work is in his own words. 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.
Is it safe to use an AI note-taker or meeting recorder for a client call transcript?
Not until the practice knows where the recording is stored, whether the 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 each question needs an answer first. The OAIC's guidance also tells organisations to review carefully the terms and settings that will apply to their use of any AI product. The OMARA says an agent needs the client's written consent before personal details go into an AI platform, and this call holds an admission against his own interest.
How can a migration practice 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 or advice email is drafted from the client file and the practice's own records. The registered migration agent checks it against what was said and signs it, because a machine transcript can mishear names, dates and numbers.
Does a call transcript satisfy the duty to record oral advice?
Section 55 of the Code of Conduct requires a contemporaneous written record of oral communications with a client, including oral instructions and advice. A transcript can form part of that record, but the agent remains responsible for it being accurate and complete. A machine transcript can mishear names, dates or numbers, so check it against what was said.
What does the Code say about advising a client whose case looks hopeless?
Section 19 bars an agent from giving assistance they reasonably believe is futile. An exception applies where a migration agent has advised the client that it is futile and the client has given a written statement that they were advised and still wish to proceed. Candid advice of this kind belongs on the client file and nowhere else.
Where does a recorded call go when an AI note-taker is used?
It depends on the tool: some process audio on the provider's servers, which may be offshore, and may keep or use the recording on the provider's terms. The OMARA says an agent needs the client's written consent before entering personal details into an AI platform, so the practice should check the tool's terms and the client's consent before recording.

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 migration practices Resources The Privacy Act, the APPs and AI All professions