AI tools are already embedded in Australian hiring processes. Automated CV screening, algorithmic candidate ranking, video interview analysis, chatbot-based initial screening, and AI-assisted reference checking are not emerging technology — they are in active use across a wide range of Australian employers, from large enterprises to mid-market businesses using off-the-shelf HR platforms. Many HR managers using these tools are doing so without a clear understanding of the legal framework that applies to them.
That framework exists. It is not a single statute written for AI — it is the intersection of existing anti-discrimination law, privacy obligations, and employment legislation applied to a new context. The absence of AI-specific regulation does not mean these tools operate in a legal vacuum. Understanding what the existing framework requires — and where regulatory attention is now focused — is the starting point for using AI in hiring responsibly.
For organisations managing significant recruitment volume through labour hire or managed workforce arrangements, how AI is used in the sourcing and screening process is a governance question that sits with both the provider and the host. Programmed’s approach to AI in our own recruitment processes is built around the principles in this article — if that’s relevant to a current evaluation, our MSP and people solutions team can walk through it directly.
Key takeaways
- Anti-discrimination law applies to the outcome of a hiring decision regardless of how it was produced — “the algorithm did it” is not a legal defence against a discrimination complaint.
- AI tools trained on historical hiring data can and do replicate past discriminatory patterns — employers are responsible for auditing the tools they use, not just the decisions they make.
- The Privacy Act creates disclosure and consent obligations when AI tools process candidate data — and the Act is being reformed to address automated decision-making specifically.
What AI in hiring actually looks like in practice
The practical landscape of AI in Australian hiring covers a wide spectrum of tools with different risk profiles:
- Automated CV screening: Software that parses applications and scores or ranks them against defined criteria — keywords, qualifications, experience patterns. The criteria are set by the employer or by the system’s training data.
- Algorithmic candidate ranking: Tools that score candidates against a model of “good hire” outcomes, typically trained on the employer’s historical hiring data. These tools learn from past patterns — including patterns that may have been discriminatory.
- Video interview analysis: Platforms that analyse recorded video interviews using facial expression analysis, speech patterns, or sentiment scoring to generate candidate assessments. This category has attracted significant scrutiny from regulators internationally.
- Chatbot screening: Automated conversational tools that conduct initial candidate qualification — asking about availability, experience, salary expectations — and either advance or eliminate candidates based on responses.
- AI-assisted reference checking: Platforms that automate referee contact, collect structured responses, and generate candidate profiles from aggregated referee input.
Each of these tools operates on candidate data, applies some form of algorithmic assessment, and produces an output that influences a hiring decision. Each of them sits within the same legal framework as a human recruiter making the same decision.
The anti-discrimination framework: outcomes, not methods
Australia’s anti-discrimination framework — the Age Discrimination Act 2004, the Disability Discrimination Act 1992, the Racial Discrimination Act 1975, the Sex Discrimination Act 1984, and state and territory equivalents — prohibits discrimination in employment on the basis of protected characteristics. The prohibition applies to outcomes: if a protected characteristic is a reason a person is treated less favourably in a hiring process, the act of discrimination has occurred regardless of how the decision was made.
This means that if an AI screening tool systematically produces lower scores for applicants over 50, or for applicants with names indicating a non-English-speaking background, or for applicants who disclose a disability — the employer using that tool is exposed to discrimination complaints. The fact that the tool produced the ranking, not a human recruiter, does not transfer the legal liability to the software vendor. The employer made the decision to use the tool and to act on its output.
Indirect discrimination is the relevant concept here. A tool doesn’t need to explicitly consider a protected characteristic to discriminate — if its outputs have a disparate impact on a protected group, and that impact isn’t justified by a genuine occupational requirement, indirect discrimination may exist. Whether a particular tool’s outputs constitute unlawful discrimination depends on specific facts and context; that analysis requires legal advice specific to the tool and the role.
The Privacy Act: disclosure, consent, and automated decisions
When a candidate submits an application, they provide personal information — name, contact details, work history, qualifications, potentially sensitive information about health or background. The Privacy Act 1988 governs how that information is collected, used, and disclosed. Feeding that information into an AI tool is a use and potentially a disclosure that must sit within the Act’s requirements.
In practice this means:
- Candidates should be informed, at the point of application, that their data will be processed by AI tools as part of the assessment process. This is not currently an explicit legal requirement in Australian law for all contexts, but it is required by the privacy principle that individuals are told how their information will be used.
- If candidate data is being sent offshore to an AI vendor’s servers — which is common with cloud-based HR platforms — the overseas disclosure requirements of the Privacy Act apply.
- Sensitive information — health, disability, criminal record — attracts heightened protections under the Act. AI tools that process this information in the screening process require careful consideration of consent and use limitations.
The Privacy Act is currently being reformed. The Attorney-General’s Department reform process has specifically identified automated decision-making in high-stakes contexts — employment is one — as an area requiring additional regulatory attention. The direction of reform is toward greater transparency obligations and, potentially, rights to human review of automated decisions. Employers building AI into hiring processes now should design for that regulatory direction, not just for the current state of the law.
Bias in AI tools: the employer’s responsibility
AI screening tools trained on historical hiring data learn to replicate the patterns in that data. If an employer’s historical hiring data reflects past patterns of underrepresentation — fewer women in technical roles, fewer workers over 50 in high-growth positions, fewer applicants from non-English-speaking backgrounds advancing past initial screening — a model trained on that data will reproduce those patterns, often invisibly.
This is not a theoretical concern. International employers — including Amazon, which publicly discontinued an AI recruitment tool that systematically downgraded resumes from women — have encountered this problem. The vendor of the tool is not the employer who bears the discrimination complaint. That is the employer’s responsibility.
Before deploying any AI screening tool, employers should seek answers to:
- What data was this tool trained on, and does that training data reflect our historical workforce or a broader dataset?
- Has the tool been audited for disparate impact on protected characteristics? By whom, using what methodology?
- What is the process for identifying and correcting bias in the tool’s outputs?
- Does the vendor provide disparate impact data for the tool’s performance across demographic groups?
Vendors who cannot answer these questions clearly are not positioned to help you manage the discrimination risk their tool creates.
What a responsible approach looks like
Responsible use of AI in hiring is not a prohibition on the technology — it is a governance framework around it:
- AI assists; humans decide. Use AI to support human judgement — to surface relevant applications, to flag inconsistencies, to reduce administrative volume — not to replace the human review of individual candidates. Final hiring decisions should involve a human who can be accountable for the basis of that decision.
- Audit tool outputs for disparate impact. Run periodic analysis of AI screening outputs against the demographic composition of your applicant pool. If the tool is systematically ranking one group lower than another, investigate before continuing to rely on it.
- Inform candidates. Update your privacy notice and application process to disclose that AI tools are used in assessment. This is emerging practice that is becoming a trust and brand expectation, not just a legal consideration.
- Maintain human review of rejections. Automated rejection of candidates at scale without any human review is the highest-risk application of AI in hiring. Build a sampling and review process into your workflow.
- Keep records. Document the basis for hiring decisions, including how AI tool outputs were used or weighted. Records are your protection in the event of a discrimination complaint.
For organisations managing workforce technology across HRIS, WFM, and VMS platforms, the governance framework for AI tools needs to be considered alongside the broader technology stack. See Workforce Technology: HRIS, WFM and VMS Explained for a practical overview. The responsible sourcing dimension — including how AI in hiring intersects with modern slavery and supply chain governance — is covered in Modern Slavery and Workforce Supply Chains in Australia.
Related reading
Also see: AI in Recruitment: Governance, Bias Risks, Safe Implementation.
This article sits within a broader cluster on advanced workforce ESG governance. For the full picture on gender pay gap reporting, modern slavery obligations, and supply chain disclosure, see Workforce ESG Reporting: Gender Pay Gap, Modern Slavery and Labour Supply Chains.
Related services
MSP and People Solutions: For organisations that want to understand how AI tools are — and are not — used in Programmed’s recruitment and screening processes, and what governance framework governs their use.
Managed Skilled Workforce: End-to-end workforce solutions with transparent sourcing and screening methodology, including documentation of the basis for placement decisions.
FAQ
If an AI tool discriminates, can we be held liable if we didn’t know?
Yes. Anti-discrimination law does not require intent — it applies to outcomes. Deploying a tool without understanding its potential for discriminatory output, and then acting on that output at scale, is not a defence. The due diligence obligation sits with the employer who chooses to use the tool. “We didn’t know the tool was biased” is not a complete answer to a discrimination complaint, though it may affect remedies and penalties.
Are there any AI hiring tools that are legally safe to use in Australia?
No tool is inherently “safe” — the question is whether the tool is used responsibly and within a governance framework that includes bias auditing, human review, candidate disclosure, and record-keeping. Tools with documented bias audits, transparent methodology, and clear vendor accountability for disparate impact create less risk than tools without those features. But the governance framework around the tool matters as much as the tool itself.
Do we need to tell candidates an AI tool reviewed their application?
There is no explicit statutory obligation in Australian law — currently — requiring disclosure of AI use in hiring assessment. However, the Privacy Act requires that individuals be informed of the purposes for which their personal information is used. Using that information as input to an AI assessment tool is a use of that information. Disclosure in your privacy notice and application process is both a legal risk-management measure and, increasingly, a candidate trust issue. Many applicants now expect to know whether AI tools have been used in their assessment.
What records should we keep when AI tools are used in a hiring decision?
Keep records of: which AI tools were used and at what stage; the outputs those tools produced for the relevant candidates; how those outputs were used in the final decision; and any human review that occurred. In the event of a discrimination complaint, these records allow you to reconstruct the basis for the decision and demonstrate that human judgement — not automated rejection — drove the outcome. Retention period should align with your privacy policy and any applicable limitations periods for discrimination claims.
Next step
If your organisation is reviewing its use of AI in recruitment and wants to understand how a responsible governance framework operates in practice, speak to Programmed’s MSP and people solutions team. We can outline how our own screening processes are governed and what that means for organisations that partner with us for managed recruitment and labour supply.
General information only: This article is for general informational purposes only and does not constitute legal advice. Legislation varies by state and territory — consult a qualified employment lawyer or Fair Work adviser for guidance specific to your situation.
