A new analysis of AI-driven hiring algorithms has found that Black and Asian job seekers are rejected at significantly higher rates than white candidates, according to reporting from The Register published this week. The findings add to a growing body of evidence that automated hiring tools — marketed as a way to reduce human bias — are scaling discrimination instead, while operating in a regulatory vacuum that leaves applicants with almost no recourse.
🔍 THE BOTTOM LINE: AI hiring tools were sold on the promise of objectivity. The reality is that they encode the biases of the data they’re trained on, run without meaningful oversight, and are now making hiring decisions at scale for thousands of companies. The discrimination isn’t accidental — it’s architectural.
❓ FAQ
Is the discrimination intentional, or an accidental byproduct of the data? Mostly accidental — but the companies deploying these tools aren’t required to check. The Register’s analysis shows the bias comes from training data that reflects historical hiring patterns. When the algorithm optimises for “people who got hired in the past,” it learns past discrimination. The opacity is the real problem: candidates have no way to know why they were rejected, and employers have no audit obligation.
Does any NZ law actually prohibit this? Not directly. The NZ Human Rights Act 1993 prohibits discrimination on race, but the burden of proof sits with the job seeker. Proving an opaque algorithm discriminated — rather than just not picking you — is a near-impossible evidentiary task. The NIST AI Risk Management Framework recommends auditing, but it’s voluntary in NZ.
What can a job seeker actually do? Practically: ask the employer whether AI screening was used, request human review, and lodge a complaint with the Human Rights Commission if you suspect racial bias. The Commission can investigate, but the process is slow and individual, not systemic. The structural fix has to come from regulators.
Will this get better as the models improve? Probably not on its own. The bias is in the data, not the model architecture. Newer, more capable models trained on the same biased historical data will produce the same biased outputs — possibly more confidently. The fix is regulatory audit requirements, not better models.
🔍 THE BOTTOM LINE
AI hiring was supposed to solve discrimination. Instead, it’s making discrimination faster, more opaque, and harder to challenge. Until regulators catch up — or until companies start auditing their own tools — the safest assumption is that the algorithm is not on your side.
What the Research Found
The analysis examined hiring outcomes from multiple AI-driven applicant tracking systems (ATS) and assessment platforms used by Fortune 500 companies, comparing rejection rates across demographic groups. While the specific platforms weren’t named in the public reporting, the pattern is consistent: Black and Asian applicants face rejection rates 20-40% higher than white candidates with comparable qualifications, even when resume content is functionally identical.
The bias enters the system in several ways:
- Training data reflects historical hiring patterns. If a company has historically hired mostly from certain universities or zip codes, the AI learns to filter for those signals — and screens out qualified candidates who don’t match.
- Language patterns get penalized. Resumes that include names associated with Black or Asian candidates, or that mention certain cultural organisations, get downranked by NLP models trained on biased corpora.
- Video interview analysis is even worse. AI tools that analyse facial expressions, speech patterns, and “enthusiasm” have been shown to systematically disadvantage non-white candidates and candidates with disabilities.
The Regulatory Gap
In the US, the Equal Employment Opportunity Commission (EEOC) has issued guidance warning that AI hiring tools can violate Title VII of the Civil Rights Act, but enforcement is almost nonexistent. In the EU, the AI Act classifies hiring AI as “high risk” and requires conformity assessments — but the rules don’t take full effect until 2027.
New Zealand sits somewhere in between. The Employment Relations Act and Human Rights Act prohibit discrimination in hiring, but there’s no specific framework for algorithmic hiring tools. The Office of the Privacy Commissioner has flagged AI hiring as a risk area, but hasn’t issued binding rules.
The result: companies can deploy AI hiring tools with almost no pre-deployment auditing, no ongoing monitoring requirements, and minimal disclosure to candidates about how decisions are being made.
The Candidate Experience
For job seekers, the impact is concrete and demoralising. A qualified candidate can be rejected by an AI before a human ever sees their resume — and the rejection often comes with no explanation, no feedback, and no appeal process. “You don’t know what you did wrong, and you can’t fix it,” one applicant told researchers. “You just get a ‘we’ve decided to move forward with other candidates’ email, and that’s it.”
This creates a particular kind of harm: discrimination that’s invisible, automated, and effectively unauditable. If a human hiring manager rejects you, you can ask why, push back, or file a complaint. If an AI does it, you’re often left with nothing to push back against.
What’s Actually Being Done
Some companies are starting to respond. A handful of large employers have commissioned third-party bias audits of their hiring AI, and some have pulled tools from production after audits revealed discriminatory patterns. The National Institute of Standards and Technology (NIST) has published a framework for evaluating AI bias, and several US states have introduced bills requiring algorithmic hiring audits.
But the audits are voluntary, the bills are fragmented, and the enforcement mechanisms are weak. Most companies using AI hiring tools are doing so with no external oversight at all.
What Job Seekers Can Do
If you’re applying for jobs in 2026, the AI hiring reality means you need to:
- Assume your first screen is automated. Tailor your resume for keyword matching, not human storytelling.
- Check if the company discloses AI use. Some employers list this in their privacy policies or job postings.
- Document everything. If you suspect discriminatory rejection, keep records — you’ll need them if you want to complain to the Human Rights Commission or EEOC.
- Apply broadly. The randomness of AI screening means applying to more roles increases your odds, even if individual applications feel like a black box.
❓ Frequently Asked Questions
Q: Can I ask a company if they’re using AI to screen my application? In some jurisdictions, yes. New Zealand doesn’t currently require disclosure, but you can ask the hiring manager directly. A refusal to answer is itself informative.
Q: Is AI hiring illegal in NZ? No — but it may breach the Human Rights Act if it has a discriminatory effect. The Privacy Commissioner has flagged AI hiring as a priority area, but there’s no specific ban or licensing requirement.
Q: What should employers do? Commission third-party bias audits before deployment, monitor outcomes across demographic groups, give candidates a human review option, and disclose AI use upfront. Anything less is asking for regulatory action.
🔍 THE BOTTOM LINE: AI hiring was supposed to solve discrimination. Instead, it’s making discrimination faster, more opaque, and harder to challenge. Until regulators catch up — or until companies start auditing their own tools — the safest assumption is that the algorithm is not on your side.
