AI governance should be part of the DEI playbook

DEI professionals have faced new challenges in 2026. Not only are their jobs continuously threatened by the Trump administrations, but many have smaller budgets. Now, with AI use seemingly ramping up in the workforce, they must also understand how to harness the technology without introducing new biases into their workflows.

On July 10, inclusion leaders met at the Meltzer Center for Diversity, Inclusion, and Belonging at New York University (NYU) School of Law to better understand how to choose and manage AI systems.

Chief among the takeaways from one panel was that while AI as a technology is underregulated in the US, equal opportunity and civil rights laws still apply.

“There is no AI fancy technology exception to our civil rights laws,” Jenny Yang, partner at Outten and Golden law firm and former EEOC commissioner, told the audience.

“Many people think an AI system designed to screen and rank applicants is actually identifying the most qualified people. Now that’s what many of these systems purport to do…but in fact, there are many fundamental concerns with these systems,” Yang said. “It’s kind of like a map. A map can help you get somewhere faster, but if you have the wrong destination, you are going to the wrong place, and that is the problem with many of these systems.”

Yang added that many current AI systems have been trained on past decisions, which may introduce biases. Instead, she recommended that DEI professionals “go back to basics” and make sure that screening systems are focused on job functions and equal opportunity principles.

Jeff McMillan, founder of McMillanAI, claimed that while there are understandable concerns about AI, it can be harnessed to monitor for bias and discrimination.

“If you are concerned about bias, have somebody build a model for you. Make sure to go through the process, and have somebody independent of that build another model, and have that model watch that model,” he said. “This technology is incredibly powerful at identifying issues, challenges. It doesn’t do a great job of predicting the future, but it’s a very good job of identifying risks and concerns.”

Emily Black, assistant professor of computer science and engineering at NYU, added that it’s important for employers to do their own research when picking a vendor because a system may be unbiased in one situation but not another.

“It’s really important to test on your own data in your own environment,” she said, adding that leaders should think about what they want to fix from their current process or what efficiencies they want from a technology. “Figure out what you want from this thing, and then talk to your IT department and find a way to measure that behavior. There are ways to measure behaviors, but we have to figure out the behaviors that we want first.”