Migration agent WhatsApp messages and AI data leakage: what a client chat thread gives away in ChatGPT or Claude.

Many migration clients prefer WhatsApp to email, so a week’s chat can hold a passport photo, a payment receipt, a health update and a confidential job search, all typed on a phone between other things. Staff paste the thread into ChatGPT or Claude to summarise it for the file or draft a reply. The Code of Conduct treats electronic communications with a client as part of the client file. This page marks every line in a routine client chat that the Code and the Privacy Act bear on, and shows how the same thread is summarised on Ermos inside the practice.

The WhatsApp thread with a client

Kept as the client messages the practice. Section 56 of the Code requires the client file to include copies of all written communications, including electronic communications, between the agent and the client. Section 53 requires documents belonging to a client to be kept securely.

WhatsApp · (placeholder)
and Kingfisher Lane Migration

Mon 26 Oct, 7.48am Joseph: Good morning Renata. (placeholder)

Mon 26 Oct, 9.05am Renata: Thanks. Please send scans through the secure portal and delete the photos from the chat.

Mon 26 Oct, 9.31am Joseph: (placeholders)

Wed 28 Oct, 6.12pm Joseph:

Wed 28 Oct, 6.40pm Renata:

Thu 29 Oct, 10.17pm Joseph:

Fri 30 Oct, 8.02am Renata: Understood, I will only message this number.

Sat 31 Oct, 8.55am Joseph:

An invented chat thread. The people and employer do not exist; the phone, passport and bank numbers are placeholders. Hover or tap any highlighted line for the analysis; the full list is below.

What leaves the practice

  1. 1
    The client’s name

    Named in the chat header. The key that ties every message to one real person. Privacy Act guide.

  2. 2
    Passport photos and a passport number

    A photo of two passports and a number typed out. Identity documents, and documents the Code requires an agent to keep securely. Privacy Act guide.

  3. 3
    A spouse’s cancer treatment

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

  4. 4
    A payment receipt with bank details

    The deposit amount and the client’s bank and account number. Financial information; in this example the bank details are placeholders. Privacy Act guide.

  5. 5
    A confidential job search

    That the client is looking for work without telling his employer, and the risk of the shared inbox. Employment information the client shared in confidence. Privacy Act guide.

  6. 6
    The agent’s advice about the health requirement

    A caution not to lodge until they talk. The practice’s advice, which the agent must keep on the client file under section 56 and stands behind; a model should not decide it.

  7. 7
    The client’s phone number

    A mobile number in the chat header. Personal information, and the account identifier for the chat. Privacy Act guide.

  8. 8
    The visa expiry

    The date the client’s visa ends. The deadline for the file. Privacy Act 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 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 WhatsApp thread with a client?
Yes, whatever is pasted or uploaded from the thread 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 thread holds two passport photos, a deposit receipt with bank details, a wife's chemotherapy and a job search the client has not told his employer about.
Is it safe to paste a WhatsApp thread with a client into ChatGPT or Claude?
No, because a chat thread puts identity, money and health into a few lines. 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. The passport number, the bank account details, the wife's chemotherapy at Westmead and the request not to be emailed at work each need protecting. 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.
What if staff use their own ChatGPT or Claude account for a WhatsApp thread with a client?
The practice cannot see or manage what was sent from a personal account. On a personal plan the user cannot choose where conversations are stored or processed. Some business plans let an organisation choose to store data at rest in Australia (OpenAI), which still leaves the content with the AI company. Both companies can also review content that is flagged for safety or policy reasons. Section 56 of the Code of Conduct puts electronic communications with the client on the client file, so a thread summarised on a personal account leaves the practice with no record of where it went.
How can a migration practice use AI for a WhatsApp thread with a client 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 summary for the file is produced from the client file and the practice's own records. The registered migration agent checks and signs it, and the advice on the health requirement remains the agent's own.
Are WhatsApp messages part of the client file?
Yes. Section 56 of the Code of Conduct requires the client file to include copies of all written communications, including electronic communications, between the agent and the client. The file must be kept for 7 years after the last action. A chat on a personal phone therefore needs to be exported to the file and not left only on the handset.
Should clients send passports by WhatsApp?
Section 53 of the Code requires documents belonging or relating to a client to be kept securely. A photo in a chat may sit on personal phones and backups. Ask clients to use the secure portal, as the agent does in this thread, and avoid pasting identity documents into any other tool.
A client asked not to be emailed at work. Does the practice record that?
Yes, as an instruction on the file, and it is exactly the kind of detail that shows why a thread should not leave the practice. Section 33 of the Code requires an agent to act in accordance with the client's instructions, and section 35 stops personal information being disclosed to a third person without written consent.

See Dominion answering from your own documents.

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

Book a demo