Lawsuits involving Eightfold AI, Meta and IBM are challenging how companies use artificial intelligence in hiring and workplace decisions. The cases allege that automated systems can rank applicants without disclosure or contribute to discrimination based on factors such as age, parental leave or medical leave.
The disputes come as employers increasingly use automation to filter applications, assess skills and conduct initial interviews. Their outcomes could influence employment practices, transparency requirements and the ability of candidates to challenge decisions made with software.
What the lawsuits allege
Erin Kistler, a product manager with nearly 20 years of experience, has applied for thousands of jobs over the past four years, including positions at PayPal, Microsoft and Netflix. She says she did not receive an interview and is suing Eightfold AI in a California class action filed in January.
Kistler’s legal team argues that Eightfold’s automated screening may function as an undisclosed consumer report or applicant dossier. The company’s software uses information from résumés, LinkedIn profiles and social media profiles to help score applicants from 0 to 5 and may prioritize some candidates for interviews. Eightfold AI denies the allegations and says it intends to defend itself.
- Workers are suing Meta over allegations that an internal AI system targeted people for layoffs after parental or medical leave.
- A lawsuit against IBM alleges discrimination against older workers. IBM denies automatically screening out candidates and says it does not tolerate discrimination.
Why researchers are concerned about bias
Researchers say automated systems can reproduce patterns in historical hiring data or introduce new forms of bias. The source material cites an earlier Amazon tool that downgraded women’s résumés after learning from a workforce whose top performers were men; Amazon discontinued the tool after the problem was found.
Research described by University of Chicago assistant professor Xuechunzi Bai found that AI models made increasingly stereotyped hiring decisions about fictional demographic groups, sometimes associating groups with occupations unrelated to qualifications. Research for Iman Ajunwa’s book also found that an AI system scored applicants with southern accents poorly during voice interviews.
- Automation can range from filtering out applicants without a four-year degree to assessing skills or conducting an initial phone interview.
- A concern among researchers is that a negative assessment could affect candidates across multiple employers using similar software or foundational AI models.
What applicants may be told
There is no US-wide law requiring employers to disclose that AI hiring systems are being used, according to Ifeoma Ajunwa, a professor at Emory University School of Law. Without access to the information used in an assessment, applicants may have no clear way to identify or dispute inaccuracies.
New York City rules that took effect in 2023 require employers using certain automated employment decision systems to conduct annual bias audits and notify candidates in advance. The requirement applies to systems that substantially assist or replace decision-making, leaving a potential gap when humans remain involved. Illinois and Colorado have also adopted laws addressing AI tools that result in unlawful discrimination.
- Some employers use AI to collect candidate information without allowing the system to automatically reject, score, rank or advance applicants.
- Incredible Health chief executive Iman Abuzeid said about 10% of the company’s AI interviews are audited for bias by a human.
No comments yet. Start the discussion.