Interview transcripts and AI privacy: what a recorded screening call gives away in ChatGPT or Claude.

Screening calls are now recorded and transcribed by an AI note-taker, and the transcript is a verbatim record of a candidate talking freely: what they earn, why they want to leave, what happened to their health, whether they plan a family and who else is interviewing them. Consultants then paste the transcript into ChatGPT or Claude to write the summary or score the candidate. This page marks every item in a routine screening transcript that the Privacy Act and anti-discrimination law touch, and shows how the same call is written up on Ermos inside the agency.

The candidate interview transcript

Recorded and transcribed on the first call with a candidate, usually 20 to 40 minutes, before a profile exists. The transcript is the raw material for the profile, the summary and the scoring.

Call transcript
Generated by an AI note-taker · Larkspur Search Partners · Tuesday 6 October 2026, 11.05am · 31 minutes

[00:00:08] Tamsin Hartigan:

[00:00:21] Nina Castellano: Yes, that is fine.

[00:03:12] Tamsin Hartigan:

[00:06:40] Nina Castellano:

[00:09:55] Nina Castellano:

[00:14:30] Nina Castellano:

[00:19:02] Nina Castellano:

[00:24:15] Nina Castellano:

[00:29:48] Tamsin Hartigan:

An invented transcript. All people and companies are fictional. Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the agency

  1. 1
    The recording and sharing script

    The words the consultant uses to tell the candidate about recording and sharing. APP 5 deals with telling an individual about collection, and the OAIC’s AI guidance links it to whether a later use was reasonably expected. A promise to share only with a client is a promise the agency has to keep, including against an AI tool. Privacy Act guide.

  2. 2
    The client and the role

    The client’s identity and the manager’s name, disclosed to a candidate on the call. Commercial information between the agency and the client.

  3. 3
    Salary, bonus and expectation

    A current package and a number the candidate would move for. Personal information, and the figure a client would pay to know. Privacy Act guide.

  4. 4
    A named manager and a job search kept secret

    A complaint about a named person and a fear of being managed out. Personal information about the candidate and about him, and the candidate’s most exposed fact. Privacy Act guide.

  5. 5
    Stress leave and her father’s stroke

    A health event for the candidate and another for her father, volunteered in conversation. Health information is sensitive information; the OAIC’s AI guidance says sensitive information generally requires consent and recommends against entering it into publicly available generative AI tools. Patient data guide.

  6. 6
    A plan to start a family

    A candidate’s intention, volunteered in answer to a question about travel. Pregnancy and family or carer’s responsibilities are protected attributes under the Fair Work Act, and a record of the plan sits in the file of a candidate who is about to be assessed. Privacy Act guide.

  7. 7
    A competing process

    Who else is interviewing her and on what money. It tells anyone who sees it what the candidate can be moved by.

  8. 8
    The consultant’s own commentary

    A candid assessment and a comment about the client’s manager, said in confidence on a recorded call.

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 agency 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 candidate 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 recruitment agencies.

Common questions

Do ChatGPT and Claude use the information in a candidate interview transcript?
Yes, everything pasted or uploaded 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. A screening transcript records pay, stress leave after a father's stroke and a plan to start a family, all in the candidate's own words.
Is it safe to paste a candidate interview transcript into ChatGPT or Claude?
No, because the consultant told the candidate the recording stays with the agency and her details go to a client only if she agrees. 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. A chatbot is neither of those, and the transcript would arrive with her salary, her health history and her family plan.
Is it safe to use an AI note-taker or meeting recorder for a candidate interview transcript?
Not without the same questions. Recording and transcription tools send the audio or transcript to the tool's vendor, so the questions are where it is stored, whether it trains a model and who can review it. 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 candidate agreed to the agency recording the call, which is a different matter from a vendor receiving it.
How can a recruitment agency use AI for a candidate interview 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. Ermos writes the summary from the transcript and the agency's own screening template. The consultant checks it against the call and signs it before it joins the candidate file.
Do we have to tell candidates we record the call?
APP 5 deals with notifying an individual about the collection of their personal information, and the OAIC links it to whether a later use was reasonably expected. Telling candidates that the call is recorded, why, and who sees it is the practical step. Check State law on recording conversations with your own adviser as well, because it is outside what this page covers.
What if the candidate volunteers health information?
Volunteered or not, it is still sensitive information once the agency holds it, and the OAIC says sensitive information generally requires consent. The practical steps are to avoid prompting for it, to keep it out of the profile and the client report unless the role needs it, and to keep the transcript out of tools you do not control.
Does an AI-generated score on a candidate create new obligations?
From 10 December 2026, APP 1.7 requires an APP entity that has arranged for a computer program to make a decision, or a step substantially and directly tied to making one, to describe it in its privacy policy where the decision could reasonably be expected to significantly affect an individual's rights or interests and the program uses their personal information.

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 recruitment agencies Resources The Privacy Act, the APPs and AI All professions