Hiring and onboarding
Using AI in recruitment: discrimination and privacy risks for Australian employers
Federal discrimination law and the Fair Work Act protect job applicants, not just employees, and a screening tool that disadvantages a group can discriminate indirectly even if nobody intended it. Applicant data is outside the Privacy Act’s employee records exemption, and from 10 December 2026 covered businesses must disclose programs that make or substantially help make significant decisions.
Updated . General information, not legal advice.
Which recruitment tools does this apply to?
Any software that screens, scores, ranks or filters applicants, including CV parsing, knock-out questions, online assessments, recorded interview analysis and AI assistants used to compare candidates.
The law doesn’t care whether the tool is called AI. What matters is whether it shapes who is offered the job. A spreadsheet formula that filters out applicants is in the same position as a machine learning model. The OAIC says AI used in decisions with a legal or similarly significant effect on someone is likely a high privacy risk activity.
Do discrimination laws protect job applicants?
Yes. The Sex, Racial, Disability and Age Discrimination Acts each make it unlawful to discriminate in deciding who should be offered employment, and the Sex, Disability and Age Acts also cover the arrangements made for deciding that.
| Act | Recruitment provision | Indirect discrimination |
|---|---|---|
| Sex Discrimination Act 1984 | Recruitment provisions 14(1): arrangements for, and determining, who should be offered employment | Indirect discriminations 5(2), with the reasonableness test in s 7B |
| Racial Discrimination Act 1975 | Recruitment provisions 15(1): refusing or failing to employ; s 15(2): people procuring employees | Indirect discriminations 9(1A) |
| Disability Discrimination Act 1992 | Recruitment provisions 15(1): arrangements for, and determining, who should be offered employment | Indirect discriminations 6, including failing to make reasonable adjustments |
| Age Discrimination Act 2004 | Recruitment provisions 18(1): arrangements for, and determining, who should be offered employment | Indirect discriminations 15 |
State and territory anti-discrimination laws also apply and may cover more attributes. The federal provisions apply to ‘an employer’, with no turnover threshold like the Privacy Act’s.
How can a screening tool discriminate indirectly?
Indirect discrimination happens when a condition, requirement or practice disadvantages people with a protected attribute and isn’t reasonable. Under the Sex, Disability and Age Discrimination Acts, the employer must prove it was reasonable.
A screening rule is a condition or requirement. If a tool learns from past hires or applies rules that track a protected attribute, it can disadvantage a group without ever seeing sex, age or disability. Examples to test for:
- penalising gaps in work history, which may disadvantage people who took parental or carer’s leave or were unwell
- rewarding a minimum number of years of experience or a graduation year, which may track age
- timed online tests or video analysis that may disadvantage people with some disabilities unless you offer adjustments
- filtering on postcode, name, language patterns or schools, which may track race or national origin.
Section 7B of the Sex Discrimination Act lists the factors for reasonableness: the nature and extent of the disadvantage, whether it could be overcome or reduced, and whether it is proportionate to the result sought. Sections 7C, 6(4) (Disability) and 15(2) (Age) put the burden of proving reasonableness on the person who imposed the requirement.
Does the Fair Work Act protect prospective employees?
Yes. Section 351 bans an employer taking adverse action against an employee or prospective employee because of attributes including race, sex, age, disability, pregnancy and family or carer’s responsibilities. Refusing to employ someone is adverse action (s 342, item 2).
Adverse action against a prospective employee means refusing to employ them, or discriminating in the terms or conditions of the job offered. Under section 361, if an applicant alleges you acted for a prohibited reason, it is presumed you did unless you prove otherwise. That is hard to do if nobody can explain why a tool rejected them.
Section 351(2) has exceptions, including action taken because of the inherent requirements of the position, and action that isn’t unlawful under anti-discrimination law where it was taken. The Fair Work Ombudsman gives examples.
Am I liable if a supplier’s tool or a recruiter discriminates?
Generally yes. Each federal Act treats acts of your employees and agents as yours (Sex Discrimination Act s 106, Racial Discrimination Act s 18A, Disability s 123, Age s 57), and the Racial, Disability and Age Acts also cover people acting on an employer’s behalf.
Under section 106 of the Sex Discrimination Act and section 18A of the Racial Discrimination Act, the defence is showing you took all reasonable steps to prevent the conduct. Ask suppliers what the tool uses as inputs, how it was tested for bias, and whether you can see results by group. Put testing and record access in the contract.
Does the Privacy Act cover job applicants’ information?
Yes, for businesses covered by the Privacy Act. The OAIC says the employee records exemption does not apply to personal information about prospective employees who are not employed, such as unsuccessful applicants.
Once someone is hired, records of their pre-employment checks become exempt employee records. Until then, and for everyone you don’t hire, the Australian Privacy Principles apply in full:
- APP 3.3: sensitive information, such as health or criminal record details, generally needs consent to collect.
- APP 5: tell applicants, at or before collection, what you collect and who it goes to, including any tool supplier.
- APP 11.2: destroy or de-identify applicant information once you no longer need it and no law requires you to keep it.
The OAIC also says the exemption is unlikely to cover organisations providing recruitment services under contract, so a recruitment platform holding your applicants’ data must meet the APPs itself. Businesses with turnover of $3 million or less are generally outside the APPs unless an exception applies (s 6D), but the discrimination laws above still apply to them.
When do the automated decision transparency rules start?
On 10 December 2026. APP 1.7 to 1.9, inserted by the Privacy and Other Legislation Amendment Act 2024, require covered businesses to explain in their privacy policy when a computer program uses personal information to make, or substantially and directly help make, significant decisions.
The commencement is set by section 2 of the amending Act: the day after 24 months from Royal Assent on 10 December 2024. The OAIC confirmed the date on 30 September 2026. The rules apply to decisions made after that date even if the tool was bought earlier.
The policy must cover the kinds of personal information used, the kinds of decisions made solely by programs, and the kinds of decisions programs substantially help make. A tool that ranks applicants for a shortlist is likely to be in scope. See automated decisions and privacy disclosure for wording.
What controls should I put around AI in hiring?
Keep a person responsible for every rejection, test outcomes by group, offer adjustments, keep records of how each decision was made, and tell candidates when software is used.
- Human review: a person with authority reviews rejections and shortlists, and can override the tool.
- Bias testing: before launch and regularly after, compare pass rates for groups you can lawfully measure, such as by age band or sex where applicants volunteer it. Investigate any gap.
- Adjustments: give applicants a way to ask for an alternative format or extra time, and act on it.
- Records: keep the job criteria, the tool’s settings, each applicant’s result and who made the final decision, so you can rebut an allegation under s 361.
- Telling candidates: say in the job ad or application form that software is used and how, and update the privacy policy before 10 December 2026.
- Supplier terms: confirm where applicant data is stored, whether it trains the supplier’s models, and how long it is kept.
The Australian Human Rights Commission’s AI and recruitment compliance checklist groups its questions into four areas: privacy and data protection, the safety and security of the system, fairness and transparency, and accountability.
Questions
Is it illegal to use AI to screen job applicants in Australia?
No. It becomes unlawful if the screening discriminates directly or indirectly because of a protected attribute without a lawful excuse, or breaches the Privacy Act for covered businesses. You remain responsible for the outcome.
Does the employee records exemption cover unsuccessful applicants?
No. The OAIC says it does not cover future employment relationships, so information about applicants who are not hired must be handled under the APPs if your business is covered by the Privacy Act.
Does a human reviewing the shortlist take the tool out of the privacy policy rules?
Not necessarily. APP 1.7 also covers programs that do something substantially and directly related to a decision, such as ranking applicants a person then chooses from.
Sources
- Sex Discrimination Act 1984 (Cth) (ss 5, 7B, 7C, 14, 106)
- Racial Discrimination Act 1975 (Cth) (ss 9, 15, 18A)
- Disability Discrimination Act 1992 (Cth) (ss 6, 15, 123)
- Age Discrimination Act 2004 (Cth) (ss 15, 18, 57)
- Fair Work Act 2009 (Cth), current compilation (ss 342, 351, 361)
- Fair Work Ombudsman: Protection from discrimination at work
- OAIC: Employee records exemption
- Privacy Act 1988 (Cth), current compilation (s 6D; APPs 3, 5, 11)
- Privacy and Other Legislation Amendment Act 2024, s 2 and Schedule 1 Part 15
- OAIC: New resources on transparency for use of AI and automated decision-making (30 September 2026)
- OAIC: Guidance on privacy and the use of commercially available AI products
- Australian Human Rights Commission: AI and recruitment compliance checklist
Checked against these sources on 12 November 2026. Spotted something out of date? Tell us.