Senate Democrats seek to establish civil rights offices for AI

Suriya Phosri/Getty Images

Sen. Ed Markey, D-Mass., is sponsoring legislation that would require all federal agencies that use artificial intelligence to set up offices “focused on bias, discrimination and other harms resulting from covered algorithms.”

A group of Democratic senators introduced legislation on Tuesday to require every federal agency that “uses, funds or oversees” artificial intelligence technologies to establish an office of civil rights to address concerns about algorithmic biases and discrimination in the emerging tools. 

The bill — the Eliminating Bias in Algorithmic Systems Act — was introduced by Sen. Ed Markey, D-Mass., and is co-sponsored by Sens. Cory Booker, D-N.J., Ben Ray Lujan, D-N.M., Jeff Merkley, D-Ore., Elizabeth Warren, D-Mass., Peter Welch, D-Vt., and Ron Wyden, D-Ore.

Markey said his legislation “will ensure that the government has the proper tools, resources and personnel to protect these communities and mitigate AI’s dangerous effects, while providing Congress with critical information to address algorithmic harms.”

The proposal would require that all relevant federal agencies create an office of civil rights and employ “experts and technologists focused on bias, discrimination and other harms resulting from covered algorithms.” 

These offices would also be required to submit periodic reports to relevant congressional committees outlining, in part, steps they have taken to address any potential consequences resulting from their uses of AI and recommendations “for legislation or administrative action to mitigate bias, discrimination and other harms from covered algorithms.”

Lawmakers, experts and activists have all cited concerns about the potential biases of AI technologies, which have been found to produce discriminatory results based on the data fed into their underlying algorithms. 

Currently, there are roughly 30 civil rights offices spread across the federal government, according to the Justice Department, although Markey said in a press release that “many are not required to secure staff with the expertise needed to advise the agency on algorithmic bias and discrimination.”

The legislative proposal has already received support from several privacy and civil rights advocacy organizations, including the Lawyers’ Committee for Civil Rights Under Law, Common Cause, the National Urban League and Public Knowledge.

“Artificial intelligence should not be used as an excuse to skirt the law, but without accountability that is what will happen,” Ishan Mehta, the director for media and democracy at Common Cause, said in a statement. “It is critical to have a functioning society that the federal government is equipped to do their job, which in the 21st century means overseeing artificial intelligence and ensuring that Americans’ civil rights are protected.”

The bill’s introduction came the same day that a landmark report released by the Government Accountability Office found that 20 non-defense federal agencies had more than 1,200 current and planned uses of AI across their systems, although the audit only identified approximately 200 instances of AI tools currently being used by the surveyed agencies.