The increasing use of artificial intelligence (“AI”) tools to assist employers with recruiting decisions invites the question of who can be held legally responsible if those decisions allegedly are discriminatory. Typically, only the employers are sued, but that may be changing.
On July 12, 2024, the U.S. District Court for the Northern District of California sustained, in relevant part, a discrimination complaint filed against Workday, Inc. by an applicant for jobs with other companies that used Workday’s AI applicant screening tools.
The plaintiff, Derek Mobley, alleged that he unsuccessfully applied for work with many Workday customers that used its AI screening tools. Mobley sued Workday for, among other things, discrimination under Title VII, the Age Discrimination in Employment Act, and the Americans with Disabilities Act. Workday moved to dismiss the amended complaint on the ground that it was not an “employer” within the meaning of those statutes and, therefore, could not be liable for any allegedly discriminatory decisions its customers made. In response, Mobley argued that he could establish Workday’s liability as the employer-customer’s “agent,” an “indirect employer,” or an “employment agency.” Notably, the Equal Employment Opportunity Commission filed an amicus brief in support of Mobley’s arguments.
The court held that Mobley plausibly alleged Workday was an “agent” of its customers and, therefore, Mobley had stated a claim for discrimination under the relevant statutes. In particular, the court highlighted his allegations that Workday’s AI screening software does not simply implement in a rote manner the criteria that employers select for hiring decisions but instead participates in the decision‑making process by recommending some candidates and rejecting others. Thus, according to the court, Mobley had plausibly alleged Workday’s customers delegated their traditional hiring function to Workday such that Workday could be liable as their “agent.”
The court distinguished other scenarios in which agency liability would not attach, such as where (i) a software vendor provides a database program to an employer, which the employer uses to filter out applicants over 40, or (ii) an email provider informs an employer’s applicants they have been rejected via email. In those cases, the court reasoned, the vendor’s software does not participate in deciding whom to hire.
Though the court allowed the case to proceed under the theory that Workday was the employer’s agent, it rejected Mobley’s argument that he plausibly alleged Workday was an “employment agency.” Finding conclusory Mobley’s allegation that “Workday’s systems source candidates,” the court held that Mobley did not plausibly allege that Workday (i) finds job opportunities for employees, such as by bringing job listings to their attention; (ii) procures employees for employers; or (iii) recruits or solicits candidates for employers. Finally, because Workday could be liable as an “agent” of its customers, there was no need to address Mobley’s alternative argument that Workday was an “indirect employer.”
Mobley will now have to present evidence to support his allegations, which will present Workday with a second opportunity to defeat his claim.
As the rate of adoption of AI technology in personnel processes accelerates, these challenges will surface with greater frequency. We can expect that the plaintiffs’ bar will continue to broaden their sights to potential defendants as they contemplate litigation. Such claims will present unique legal questions as to both the companies developing these AI tools and the companies using them. The court’s decision reinforces the criticality of interpretability in AI used for decisions about employment opportunities.
