Artificial intelligence is rapidly reshaping hiring. Employers increasingly use AI-enabled recruiting tools to screen resumes, rank applicants, administer assessments, and support hiring decisions. These tools may improve efficiency and consistency while also creating new litigation risk under existing employment discrimination laws and newly developing regulations.
Implementing AI tools in the hiring process warrants careful assessment of evolving AI-specific state and local regulations, including notice requirements, potential opt-out rights, and requirements for bias audits. Because some of these laws do not expressly create a private right of action, enforcement is often left principally to government agencies.
Beyond these emerging AI requirements, employers may wish to prepare for litigation challenging AI-assisted hiring decisions. Plaintiffs’ attorneys are increasingly alleging these tools illegally discriminate under traditional discrimination theories, and courts are confronting novel questions about discovery, algorithmic transparency, and employer responsibility for vendor technology.
The technology may be new; the legal principles are not. This GT Alert explores those legal principles, impact on discovery, and the role governance plays for risk mitigation.
Existing Disparate Impact Principles Apply to AI
Title VII expressly recognizes disparate-impact claims based on employment practices that disproportionately exclude members of protected classes. Under a disparate impact theory, a facially neutral policy or practice may be unlawful if it disproportionately affects members of a protected group, even absent discriminatory intent, unless the employer can satisfy a statutory defense, such as that the practice is job related and consistent with business necessity. Employers may not assume that automation changes the legal standard or shields hiring decisions from challenge.
Adding to the legal landscape are recent changes during the Trump administration, including, but not limited to:
- On April 23, 2025, President Trump issued an executive order entitled “Restoring Equality of Opportunity and Meritocracy” that aims to “eliminate the use of disparate-impact liability in all contexts to the maximum degree possible.” Read more in our GT Alert.
- It was widely reported that on Sept. 15, 2025, the EEOC distributed an internal memo that directed the agency to discharge all disparate impact discrimination claims and align with the April 23 executive order, which further instructed federal agencies to deprioritize disparate impact claims.
- On June 9, 2026, the Justice Department Office of Legal Counsel issued a formal opinion that the EEOC disparate impact guidelines violate the constitution because they “pressured employers to engage in racial discrimination.”
In response to the evolving federal landscape, some states, including California, Illinois, and New Jersey, have codified disparate impact liability under the state anti-discrimination laws. For example, the Illinois Human Rights Act now provides an express state law basis for challenging facially neutral employment practices that disproportionately affect protected groups.
AI systems can manifest disparate impact in familiar ways. Bias might arise where an AI system produces outputs that are biased or prejudiced against protected characteristics, such as gender, age, race, and disability. Three key types of potential bias are: human bias, algorithm bias, and data bias. Human biases may affect the data produced and collected and might be unknowingly fed into the algorithms we create. The potential for algorithm bias arises when an algorithm produces results that are systematically prejudiced due to assumptions in the machine learning process. Data bias occurs where the AI tool has been trained on incomplete or imbalanced data that is not representative of the general population. Thus, tools trained on historical workforce data may replicate or amplify existing imbalances. Even if protected characteristics such as race or sex are excluded, neutral variables like education, work history, geography, or language patterns may inadvertently operate as proxies.
Discovery May Look Different in AI Hiring Litigation
AI hiring cases may also change the discovery landscape. Traditional discrimination cases often focus on the human decision-maker and the reasons for the challenged decision. When AI screens or ranks applicants, litigants might seek discovery concerning the design, development, deployment, and operation of the tool, including model inputs and outputs; training and validation data; selection, scoring, ranking, and weighting criteria; adverse impact and bias testing; validation studies; audit reports; system updates and change logs; vendor communications and contracts; policies, training, and practices governing the tool’s use; records reflecting human review or overrides of AI-generated recommendations; complaints or concerns regarding the tool’s performance; and documents showing the employer’s evaluation of the tool’s accuracy, fairness, and compliance with applicable equal employment opportunity laws. Employers may not have direct access to critical defense materials because key technical documentation may remain with the vendor, raising issues of possession, custody, control, confidentiality, and trade-secret protection.
Preservation may also pose additional challenges. AI systems may change through updates, retraining, or model revisions. When litigation arises, employers should consider preserving information related to the version of the AI hiring tool used during the relevant hiring period, along with applicant scoring data, audit logs, validation studies, bias-testing results, system change logs, and records reflecting any human review of AI-generated recommendations. Because these materials may become central to claims asserting disparate treatment or disparate impact, employers may wish to account for AI-related evidence in their litigation hold and document preservation processes.
Governance Remains a Key Risk Mitigation Tool
As AI-specific employment regulation continues to evolve, practical risk mitigation involves familiar strategies updated for AI technology. Employers using AI-assisted hiring tools should consider:
- conducting diligence before implementing an AI system;
- validating the tool for its intended hiring purpose;
- providing training to staff on how to responsibly use the AI tool;
- documenting governance procedures and review processes;
- maintaining meaningful human oversight;
- monitoring hiring outcomes for potential adverse impact;
- partnering with legal counsel to conduct bias and fairness assessments; and
- planning for preservation of AI-related information and coordination with vendors.
Risk mitigation may not fully eliminate litigation risk. But targeted proactive measures might help employers show that AI-assisted hiring decisions serve legitimate business objectives and are subject to appropriate oversight. As AI becomes more embedded in recruiting, employers might experience closer scrutiny from regulators and private litigants. Key issues may focus on how AI tools were implemented, whether they produced disparate outcomes, and whether the employer can explain and defend the decision-making process. Employers who evaluate their AI hiring tools, document their processes, and monitor outcomes may be better positioned to capture the benefits of AI in hiring while reducing litigation risk.