Examples of AI Bias in Recruitment: Cases and Lessons for Employers

The risks of AI hiring bias are not hypothetical. A small number of well-documented cases and research findings have shaped regulation and practice. Each carries a specific lesson.

By the HRight Talks editorial teamUpdated 25 September 20264 minute read

Short answer

Well-known examples of AI bias in recruitment include Amazon's experimental CV screening model, abandoned after it learned to penalise references to women, as reported by Reuters in 2018; the US EEOC's 2023 settlement with iTutorGroup over application software that automatically rejected older applicants; research showing facial analysis and speech recognition systems perform less accurately for some demographic groups; and litigation testing whether AI screening vendors can share liability with employers.

Key takeaways

Bias learned from history is the most common and most predictable failure.
Hard-coded rules can be just as discriminatory as machine learning.
Technologies that analyse faces, voices or speech carry additional risk of uneven accuracy.
Liability can extend beyond the employer to vendors, depending on jurisdiction.

Case 1: Amazon's experimental recruiting model

What happened: Reuters reported in 2018 that Amazon had built an experimental tool to rate job candidates, trained on CVs submitted over about ten years. Because most came from men, the model learned to downgrade CVs containing the word "women's", as in "women's chess club captain", and graduates of certain all-women's colleges. Amazon said the tool was never used by its recruiters to evaluate candidates and abandoned the project.

Lesson: training a model to imitate past hiring decisions reproduces the biases in those decisions. Test for adverse impact before use, and avoid ranking models trained solely on historical outcomes.

Case 2: Automatic age-based rejection

What happened: In 2023 the US Equal Employment Opportunity Commission settled its first lawsuit involving AI-related hiring discrimination, against iTutorGroup. The EEOC alleged the company's application software was programmed to automatically reject female applicants aged 55 or older and male applicants aged 60 or older.

Lesson: discrimination does not require sophisticated machine learning. Simple automated rules can be unlawful, and automation removes the human who might have noticed. Review every knockout rule for legal and job relevance.

Case 3: Uneven accuracy in facial analysis

What happened: The 2018 Gender Shades study by Joy Buolamwini and Timnit Gebru found that commercial facial analysis systems misclassified the gender of darker-skinned women far more often than lighter-skinned men. Concerns about facial and emotion analysis in hiring grew, and at least one major interview platform announced in 2021 that it had stopped using facial analysis in its assessments.

Lesson: avoid technologies that infer characteristics from faces or expressions. The EU AI Act now prohibits AI that infers emotions in the workplace, except for medical or safety reasons.

Case 4: Speech recognition gaps

What happened: A 2020 study published in the Proceedings of the National Academy of Sciences found that leading automated speech recognition systems made substantially more errors transcribing Black speakers than white speakers.

Lesson: if AI transcribes and scores spoken interviews, transcription errors can translate into lower scores for some groups, and for candidates with speech-related disabilities or strong regional accents. Test accuracy across your applicant population. See is AI interview assessment fair?

Case 5: Vendor liability litigation

What happened: In the United States, Mobley v. Workday alleges that AI-enabled applicant screening in a widely used HR platform discriminated against applicants. In 2024 a federal court allowed claims to proceed on the theory that a software vendor could be liable as an agent of the employers using its tools. The case later proceeded as a collective action on age claims, and in 2026 the court rejected the argument that age discrimination law does not protect job applicants. In June 2026 it largely denied a further motion to dismiss, keeping claims based on race, sex, age and disability moving, including claims under California state law. The case was ongoing at the time of writing.

Lesson: both employers and vendors face scrutiny. Contracts should address bias testing, audit cooperation and responsibilities, and employers should not assume a vendor's reputation removes their own duties.

Common threads

Failure pattern	Cases	Prevention
Learning from biased history	Amazon	Job-analysis criteria; adverse impact testing
Discriminatory hard-coded rules	iTutorGroup	Legal review of every knockout rule
Uneven technical accuracy	Facial analysis, speech recognition	Avoid inference from faces; test accuracy across groups
Unclear accountability	Vendor liability litigation	Contractual allocation; employer oversight

See how to audit AI hiring tools and how to reduce bias in AI recruitment.

Case summaries are based on public reporting and official statements; ongoing legal matters may have developed since publication. This is general information, not legal advice. Anti-discrimination and AI rules differ by country and state; take qualified advice before relying on any tool for decisions about people.

AI Hiring Bias: Causes, Real Cases, Law and How to Prevent It
Where AI hiring bias comes from, how it is measured, the law, and a prevention framework.
AI in HR Examples and Use Cases: 10 Scenarios Across the HR Function
Ten detailed scenarios showing how AI works in practice, plus lessons from public successes and failures.
Is AI Interview Assessment Fair? Evidence, Risks and Safeguards
Where AI interviews can be fairer, where they are not, and the safeguards that make the difference.
Ethical Issues of AI in Human Resources: 10 Dilemmas HR Must Navigate
Ten ethical dilemmas of AI in HR and the questions to ask about each.

Frequently asked questions

What is the most famous example of AI hiring bias?

Amazon's experimental recruiting model, reported by Reuters in 2018, which learned from a decade of male-dominated hiring data to downgrade CVs mentioning women's activities. Amazon abandoned it.

Has anyone been sued for AI hiring discrimination?

Yes. The US EEOC settled a case with iTutorGroup in 2023 over software that automatically rejected older applicants, and litigation against HR technology vendors over AI screening is ongoing in the United States.

Why is facial analysis risky in hiring?

Research has shown facial analysis can be less accurate for some demographic groups, and inferring traits or emotions from faces has weak scientific support. The EU AI Act prohibits emotion recognition in the workplace except for medical or safety reasons.

What do AI bias cases have in common?

Most involve learning from biased historical data, discriminatory automated rules, uneven accuracy across groups or unclear accountability, and most could have been caught by testing outcomes before and after deployment.