Experts caution against AI discrimination at NRF Retail Law Summit

Retailers need to be sure they do not use artificial intelligence in ways that could result in unlawful discrimination as adoption of the popular technology expands and the Trump administration seeks to make the United States a global leader in AI. 

That was the message from legal experts and officials from the Equal Opportunity Employment Commission and Labor Department who spoke earlier this month at NRF’s annual Retail Law Summit. 

Adam S. Forman from Epstein Becker Green moderated a session on “AI, Law and Leadership: Steering Workplace Strategy Through Political Change.” He was joined by EBG’s Nathaniel Glasser, with whom he co-chairs the law firm’s AI working group, EEOC Commissioner Brittany Bull Panuccio and Labor Department Chief Innovation Officer Taylor Stockton. 

The session was one of several top-level discussions at the three-day online event, which also covered topics ranging from data security to product stewardship and drew record attendance of over 1,600. Participants were eligible for continuing legal education and Society of Human Resource Management credits. 

Shifts in federal policy 

Over the past year, President Donald Trump has signed a series of executive orders intended to move federal AI policy away from the approach taken by the Biden administration that focused on potential risks and instead prioritize deregulation and preemption of state and local AI regulations. 

Under those orders, the EEOC and other federal agencies have been directed to remove diversity, equity and inclusion concepts from their AI framework, Forman said. In addition, the EEOC has been told to “deprioritize” investigation of broad claims that artificial intelligence has a disparate impact based on legally protected status such as age, race or gender and instead focus on specific cases of discrimination, Panuccio said. 

Panuccio said disparate treatment “remains unlawful” and that the changes don’t mean the EEOC won’t take action against AI-related discrimination. In one recent case, the commission settled a claim involving a company that used AI software that automatically rejected female job applicants over age 55 and male applicants over 60 in violation of the Age Discrimination in Employment Act, she said. Other claims have involved AI technology that couldn’t accommodate job applicants unable to speak English or those with disabilities covered under the Americans with Disabilities Act. 

“I want to emphasize that the EEOC continues to vigorously enforce workplace antidiscrimination laws, including in cases implicating AI,” Panuccio said. “I’m personally invested in ensuring that AI isn’t a barrier to applicants or employees based on their protected status.” 

Practical implications for employers 

Even if the EEOC has changed its approach, Forman said employers could still face lawsuits over disparate impact discrimination. 

“The fact that the Commission is not going to be investigating or prosecuting disparate impact may lead some to think they need not think about it, but it’s still part of the statute,” he said. “These are still these people out there called plaintiff’s attorneys.” 

Glasser agreed. 

“It’s still a recognized cause of action under several federal civil rights statutes,” he said. “We would anticipate that the plaintiff’s bar will continue to pursue those theories in court.” 

Similar to incidents cited by Panuccio, some lawsuits have focused on artificial intelligence that sets criteria like age cutoffs or doesn’t accommodate language and disability needs, Glasser said. One argued that using AI to rank how job applicants’ qualifications compare with job requirements amounts to acting as a consumer reporting agency as defined under the Fair Credit Reporting Act and is therefore subject to regulation. 

Employers are responsible for the actions of both in-house and third-party AI tools they use, so it is important “that you’re not signing onto a product that is intentionally weeding out or selecting applicants or employees based on a protected characteristic,” Glasser said. 

While most AI discrimination concerns focus on recruiting, hiring and promotion, Forman said AI could be considered an agent of a company, particularly AI chatbots. A lawsuit could argue that “communication that an individual has with a chatbot may be cited as evidence that the company knew of a certain problem,” he said. 

Citing the example of an employee raising a harassment complaint via a chatbot, he said companies should make sure their policies clearly identify human contacts with whom workers can file a complaint and specifically exclude chatbots. 

The orders signed by Trump seek to preempt state and local AI laws, a move Forman said could help retail companies operating in multiple states or nationally. Local artificial intelligence measures have been adopted in California, Colorado, Illinois and New York City and are being considered elsewhere. Glasser said the executive orders don’t have the authority to preempt those initiatives, but the Department of Justice is creating a task force to challenge them and the move “signals a significant push toward federal preemption.” 

At the Labor Department, Stockton said officials are working to dispel the idea that AI will put millions of Americans out of work. 

“In all cases of every past wave of technology, and what we expect to see for this wave, is that many more jobs will be created than lost,” he said. AI will relieve workers of routine tasks by letting them work more efficiently, and there are still “messy tasks” that require human involvement. 

Stockton said the department is “prioritizing AI literacy” to ensure workers can take advantage of new jobs created by AI. In addition, the department is building a “talent pipeline” for workers needed to build and operate AI infrastructure such as data centers.