Bill introduced in U.S. Congress to regulate machine learning algorithms
U.S. congressional democrats have introduced a bill in both houses to force large companies to audit machine learning algorithms, such as for facial biometrics, for bias, The Verge reports.
The Algorithmic Accountability Act (PDF) is being sponsored in the Senate by Cory Brooker (D-NJ) and Ron Wyden (D-OR), and in the House by Representative Yvette Clarke (D-NY). It gives the Federal Trade Commission a mandate to create rules for evaluating automated systems deemed “highly sensitive,” including their training data, and require companies to perform a self-assessment based on the criteria. The company is then expected to take corrective action if an algorithm is found to pose a risk of discrimination or privacy loss.
The bill only applies to data brokers and related businesses and to companies that meet thresholds of $50 million in earnings, or possession of data on either a million customers or customer devices.
The announcement notes that Facebook was charged with violations of the Fair Housing Act by the Department of Housing and Urban Development earlier this month, and refers to reports of an automated recruiting tool that Amazon shut down after discovering it was biased against women. Algorithmic accountability has been a hot topic in the biometrics industry over the past year, with the Algorithmic Justice League launching a Safe Face Pledge campaign in late 2018, and Senator Kamala Harris calling on regulators to specifically consider facial recognition in examining AI bias last September.
The Center of Privacy and Technology at Georgetown Law, which has been critical of the use of facial recognition for CBP’s Biometric Exit program, has endorsed the bill, along with Data for Black Lives and the National Hispanic Media Coalition.
“By requiring large companies to not turn a blind eye towards unintended impacts of their automated systems, the Algorithmic Accountability Act ensures 21st Century technologies are tools of empowerment, rather than marginalization, while also bolstering the security and privacy of all consumers,” Clarke says in a statement.
New York City passed an algorithmic transparency bill in 2017, and Washington State has considered a similar measure this year, according to The Verge.