Recruiter Graduate Application Forms: Use case and AI Confidentiality
A graduate intake produces hundreds of application forms in a fortnight, which is why agencies reach for ChatGPT or Claude to score, rank and summarise them. Each form is a young person’s first real file: birth date, marks, background, the adjustments they ask for, the clubs they belong to and a tick-box consent they barely read. This page marks every item in a routine graduate application that the Privacy Act and anti-discrimination law touch, including the privacy policy statements about automated decisions that start on 10 December 2026, and shows how the same form is handled on Ermos inside the agency.
The graduate application form
Completed online by each applicant before the closing date and loaded into the agency’s system, where it is scored and ranked before any shortlist goes to the client. The office-use block and consultant comments are added by the agency afterwards.
Graduate Pathways 2027 · Application
Larkspur Search Partners for Ingleby Building Supplies · Received Sunday 8 November 2026
| Programme | Graduate Pathways 2027. Two places, twelve-month rotation, start Monday 1 February 2027. Applications close Friday 20 November 2026.1 |
| Applicant | Isla Hutchens, born 3 March 2004, 14 Cobar Street, Greenacre NSW 2190, 04xx xxx 418, isla.hutchens@example.com2 |
| Study | Bachelor of Commerce (Accounting and Supply Chain), a Sydney university, completes December 2026. Weighted average mark 71.4. First in her family at university.3 |
| Background (optional) | Born in Colombo, arrived 2011, Australian citizen. Language at home: Sinhala. Does not identify as Aboriginal or Torres Strait Islander.4 |
| Work and caring | Pharmacy assistant, 14 hours a week. Cares for her mother, who has multiple sclerosis, on weekends and may need Thursday afternoons for her mother’s infusion appointments.5 |
| Adjustments (optional) | Diagnosed with ADHD at 19. Would like extra time and a quiet room for the online assessment.6 |
| Activities | Treasurer, the university Labor club, 2025 to 2026. Volunteer at a community legal centre.7 |
| Consent | Ticked 8 November 2026: I agree that Larkspur may use software to score and rank my application, and may share my details with Ingleby and with technology providers, some of which store data overseas.8 |
| Office use | Auto-screen score 58 of 100. Cut-off 60. Status: not progressed to client. Reviewed by: no one.9 |
| Consultant comment | DK: with the ADHD and the caring she will struggle with Ingleby’s Saturday roster. The score would have parked her anyway.10 |
An invented application. The applicant, the client and the phone number and email address are fictional or placeholders. Hover or tap any highlighted item for the analysis; the full list is below.
Graduate application summariser
An Agent+ that reads a graduate application and sets out the evidence against the selection criteria your agency has agreed with the client. It leaves out details such as birth date and personal circumstances, and flags incomplete answers, so a recruiter builds the shortlist from a consistent summary.
What it does
- Reads the application and the client's selection criteria
- Summarises the evidence against each criterion
- Flags incomplete answers and unrelated personal details
What stays with your people. A recruiter reviews every summary and decides who is shortlisted. The agent does not score, rank or reject applicants.
An Agent+ is built once and anyone in the practice can run it. It works from your own documents and runs on Ermos Dominion, the AI unit in your office, or on Ermos Edge, your own instance in an Australian data centre.
What leaves the agency
- 1The programme details
The number of places, the rotation, the start date and the closing date. The same facts appear on the client’s careers page. Nothing in them is private.
Frontier Models (ChatGPT, Claude) Public details; harmless to send Ermos Edge Not about where data goes Ermos Dominion Not about where data goes Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 2Identity and contact details
A named young adult with a birth date, a home address, a mobile number and an email address. Personal information, and the set of facts used to verify identity. APP 5.1 requires the agency to take such steps (if any) as are reasonable, at or before collection or as soon as practicable after, to notify the applicant of the matters in APP 5.2 or otherwise to ensure the applicant is aware of them. Privacy Act guide.
Frontier Models (ChatGPT, Claude) Transmitted to the provider, 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 Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 3Marks, degree and first-in-family status
Results and study details collected to assess a graduate. APP 3.2 limits collection of personal information to what is reasonably necessary for the organisation’s functions or activities. First-in-family status is something the applicant chose to give, and it shapes how the agency may use it. Privacy Act guide.
Frontier Models (ChatGPT, Claude) Marks and study details 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 Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 4Country of birth, language at home and Indigenous identity
Optional background questions. Information about racial or ethnic origin is sensitive information under the Privacy Act, and country of birth, language at home and Indigenous identity can each point to it. APP 3.3 generally requires consent and that the collection be reasonably necessary for a function or activity. Race, colour and national extraction are attributes in section 351(1) of the Fair Work Act, and in the NSW Act race includes nationality, descent and ethnic origin. Optional diversity answers need a stated purpose and should be kept away from whoever ranks the application. Privacy Act guide.
Frontier Models (ChatGPT, Claude) Sensitive information transmitted; on consumer tiers it can train future models Ermos Edge Stays on your instance; a permitted excerpt may go to the named Australian provider, which keeps nothing Ermos Dominion Stays on the unit in your office Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 5A mother’s illness and the applicant’s caring role
Health information about a third person, with nothing to show she has consented to its collection (APP 3.3 generally needs the consent of the person the information is about), and a statement about the applicant’s caring role. Information about a person’s illness is health information, which is sensitive information. Family or carer’s responsibilities are an attribute in section 351(1) of the Fair Work Act, and Part 4B of the NSW Act deals with responsibilities as a carer. Patient data guide.
Frontier Models (ChatGPT, Claude) A third person’s illness and a caring role transmitted; health information is sensitive information Ermos Edge Stays on your instance; a permitted excerpt may go to the named Australian provider, which keeps nothing Ermos Dominion Stays on the unit in your office Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 6An ADHD diagnosis and a request for adjustments
Health information includes information or an opinion about the health, including an illness, disability or injury, of an individual, and health information is sensitive information. Physical or mental disability is an attribute in section 351(1) of the Fair Work Act, and Part 4A of the NSW Act and the Disability Discrimination Act 1992 (Cth), whose section 21 binds employment agencies, deal with disability. Asking what adjustment someone needs for an assessment is narrower than asking about a condition; collect only what the adjustment requires, with consent, and keep it away from the ranking. Patient data guide.
Frontier Models (ChatGPT, Claude) Sensitive information transmitted; on consumer tiers it can train future models Ermos Edge Stays on your instance; a permitted excerpt may go to the named Australian provider, which keeps nothing Ermos Dominion Stays on the unit in your office Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 7A university political club
Political opinions and membership of a political association are both on the Privacy Act list of sensitive information, and political opinion is an attribute in section 351(1) of the Fair Work Act. Once the answer is in the agency’s file it is sensitive information the agency holds. Privacy Act guide.
Frontier Models (ChatGPT, Claude) Political association transmitted; it is sensitive information Ermos Edge Stays on your instance; a permitted excerpt may go to the named Australian provider, which keeps nothing Ermos Dominion Stays on the unit in your office Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 8One tick covering software scoring, sharing and overseas storage
A single consent for three different things, naming no provider and no country. APP 5.2(i) and (j) are about whether the entity is likely to disclose personal information to overseas recipients and, if practicable, which countries. Consent under the Act means express or implied consent, and whether a bundled tick amounts to a meaningful one is a question for the agency. The OAIC recommends as best practice that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools. Privacy Act guide.
Frontier Models (ChatGPT, Claude) The consent covers the agency’s own software and providers rather than a chatbot a staff member picked Ermos Edge The notice can name the Australian provider; the wording of the consent is still the agency’s job Ermos Dominion Stays on the unit; the notice still has to describe how the agency scores applications Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 9An automated score and a cut-off with no reviewer
A score that decided the outcome, with nobody reviewing it. From 10 December 2026 APP 1.7 to 1.9 add content to the APP privacy policy of an APP entity that has arranged for a computer program to make, or do a thing substantially and directly related to making, a decision that could reasonably be expected to significantly affect the rights or interests of an individual, using personal information about that individual. The amendment applies to decisions made after that date, so an application scored in November 2026 falls before it, while the same screening after 10 December 2026 could be caught. These provisions concern what the privacy policy says; they do not themselves require a human reviewer, and whether a screening cut-off meets the test is for the agency to assess. APP 10.1 requires reasonable steps to ensure the personal information collected is accurate, up to date and complete, and the OAIC’s recruiting example calls for controls so outputs are accurate and not biased, with human oversight. Privacy Act guide.
Frontier Models (ChatGPT, Claude) The score and the cut-off transmitted; sending them does not make the screening sound Ermos Edge Ermos keeps the data in; any policy statement and human oversight of the screening are still the agency’s job Ermos Dominion Ermos keeps the data in; any policy statement and human oversight of the screening are still the agency’s job Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case - 10A consultant’s comment linking illness and caring to a roster
A candid assumption about what the applicant can manage, based on her health condition and her caring role, recorded before any assessment. Disability and family or carer’s responsibilities are attributes in section 351(1), and under section 361 a prohibited reason is presumed if alleged, unless the person proves otherwise. If the applicant asks for what the agency holds, APP 12.1 requires access unless an exception in APP 12.3 applies; paragraph (j) concerns evaluative information generated within the entity in connection with a commercially sensitive decision-making process, and whether it applies is for the agency. Privacy Act guide.
Frontier Models (ChatGPT, Claude) A candid view linking illness to a roster transmitted Ermos Edge Stays on your instance; what is written about an applicant is still a person’s decision Ermos Dominion Stays on the unit; what is written about an applicant is still a person’s decision Use case Graduate application summariser. Summarises each application against the criteria the client set, leaving out personal details, so a recruiter can review it before any shortlist is built. See the use case
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 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 Use cases in recruitment agencies.
Common questions
Do ChatGPT and Claude use the information in a graduate application form?
Can we use ChatGPT or Claude for a graduate application form if we remove the applicant's name?
Can I upload a graduate application form to ChatGPT or Claude instead of pasting it?
How can a recruitment agency use AI for a graduate application form without sending it to ChatGPT or Claude?
When do the automated decision rules in the Privacy Act start?
Is disability information on a graduate application form sensitive information?
Can an applicant ask to see their screening score?
See Dominion answering from your own documents.
A demonstration takes 30 minutes, uses no client data, and comes with no obligation.
Book a demo