In the News

Black leaders in Congress say that they will use the power of their numbers to force civil rights protections in any artificial intelligence bill that is about to become law.

The Congressional Black Caucus controls a key bloc of votes in Congress, with close to 60 Democratic lawmakers in its ranks. When civil rights provisions were removed from House lawmakers’ attempt at landmark privacy legislation earlier this summer, the bipartisan legislation stalled, in part, because of opposition from Black members. 

“Because of the large bloc of votes that we deliver, there’s nothing that’s going to move without us,” said Rep. Steven Horsford (D-Nev.), the caucus’s chairman.

Congress has yet to pass major AI legislation, but the technology is evolving rapidly, and caucus members are highlighting the dangers and promises of the new technology where and when they can.

“I don’t think any of us have not heard a story out there amongst our constituents where someone hasn’t been harmed in some form or fashion based on flawed data sets,” said Rep. Yvette Clarke (D-N.Y.), a vice chair of the caucus, which for decades has acted as a sentinel looking out for discrimination. 

This week, members and allies will highlight their concerns as well as the economic opportunities they see from the rapid development of AI at the annual legislative conference of the Congressional Black Caucus Foundation, which begins Wednesday in Washington. Topics include the impact of AI in healthcare, employment, housing and voting information. Sen. Laphonza Butler (D-Calif.) will speak about how AI can empower black communities. 

Disproportionate Bias

The caucus is developing a multi-pronged approach to AI: advocating for federal legislation; regulation, which so far has been driven largely by President Joe Biden’s executive order on AI; and state laws, where lawmakers in some states have been much more active than on Capitol Hill. Major AI bills are pending at the desks of California Gov. Gavin Newsom (D) and New York Gov. Kathy Hochul(D). 

Horsford also stresses the economic opportunities.

“We just need to be at the forefront so that we ensure that we’re not excluded like other sectors have done to black and marginalized communities,” he said.

In June, Horsford, Rep. Maxine Waters (D-Calif.), and other lawmakers went to Silicon Valley in California to meet with representatives of Apple Inc., Open AIand researchers at the Berkeley Artificial Intelligence Research lab. The caucus organized briefings that included representatives from Meta and Microsoft.

They have engaged with top AI researchers at Stanford University in California, where in March the university’s Institute for Human-Centered Artificial Intelligence released a paper titled “Exploring the Impact of AI on Black Americans: Considerations for the Congressional Black Caucus’s Policy Initiatives” after conversations with the caucus.

“We write this white paper with the conviction that the CBC has more to contribute to AI policy than simply correcting racial biases,” the paper said. “Instead, it can help steer AI to ensure the well-being and prosperity of Black communities.”

The paper pointed out research on how tools to detect deepfakes perform best when working with White faces, how automation might affect Black Americans more due to higher numbers of the population that work jobs such as truck drivers or in the food service industry, and research showing evidence of racial bias in a predictive tools used by health insurance companies.

A Bloomberg analysis last year of more than 5,000 images generated by Stable Diffusion—an AI text-to-image generator that produces images from written prompts—found that it amplified racial and gender stereotypes. Images of individuals generated for high-paying jobs were dominated by those with lighter skin tones while those generated by queries such as “fast food worker” and “social worker” usually had darker skin tones.

Russell Wald, deputy director of Stanford HAI, said that AI models can have a “unique and disproportionate” bias toward people of color. 

“We see that in the data,” he said. “The issue is how do you mitigate. How do you work in this context and understand the needs of people of color?” 

Sanmi Koyejo, a Stanford computer science professor who is also the president of Black in AI, an organization whose aim is to increase the presence of Black people in the AI field, said that the goal was to engage lawmakers early, for better outcomes.

“There is a general feeling that the gap between humans and machines is that humans are somehow more likely to be irrational, and machines and algorithms have this benefit of by virtue of being algorithmic, seem more rational,” Koyejo said. “People’s decisions end up shaping what these algorithms are and what the impacts are.”

‘Make Some Difference’

Horsford said he has been mulling what pressures the caucus can bring to bear. “Where can you make some movement? Where can you make some difference?”

His Nevada district includes large numbers of workers in the hospitality and entertainment industries, two areas where AI is being used extensively. “I don’t think that we’ve seen the worst of what it can be yet,” he said.

Back in Washington, Horsford said he has concluded AI technologies need to be deployed with civil rights protections at the center. And that is part of a policy recommendation that the caucus will likely make in the coming months.

“This is not a policy series where a set of recommendations will be made and not acted upon,” he said in late August. “We are prepared in the coming months to include provisions in the overall House Democratic platform of bills.”