Introduced by California Rep. Mark Takano, the bill would amend the Federal Rules of Evidence so that proprietary algorithm owners can’t use the trade secret privilege to avoid sharing information about their program with defendants. It also would guarantee defendants access to a working software version, with the data needed to reproduce the results presented in court, and create a standards and testing program for forensic algorithms.
The spark for the bill came from Wexler’s 2015 Slate op-ed “Convicted by Code,” which outlined how some criminal defendants were being denied access to the guts of programs used to make a DNA match or pinpoint their location. Algorithm owners and developers, which sometimes are governmental entities, have successfully argued that code details are trade secrets that must be shielded from competitors—and from defense attorneys.
“It just struck me as wrong,” says Wexler, who in 2018 published another paper expanding her argument and asserting that the trade secret privilege should not stop disclosure in criminal cases.
Errors in algorithms—or inadvertent coding that can amplify racial and gender bias—are well-documented. For Wexler, it’s also vital that defendants can access the trade secret in question as they defend themselves.
“That’s something the law is doing for businesses and innovators,” she says. “Why can’t the law also do that for criminal defendants, when the stakes can be someone’s life or liberty?”
Algorithms are used throughout criminal proceedings, from gathering evidence to sentencing and parole recommendations, and the science holds great promise for improving accuracy. But as a wave of exonerations based on DNA evidence has shown, forensic evidence isn’t foolproof, Roth points out.
“The level of sophistication of proprietary algorithms now used to generate proof of guilt is unprecedented,” says Roth, who helped Takano’s staff incorporate the idea of standards into the bill and whose research explores how science-based prosecutions affect evidence law. “It’s something that existing rules of evidence aren’t fully equipped to deal with.”
Wexler and Roth, both faculty codirectors of the Berkeley Center for Law & Technology, have each cited a California Law Review student note on computer code, intellectual property, and criminal prosecutions written by Christian Chessman ’18. They also were influenced by the work of Clinical Professor Jennifer Urban ’00, who has advocated for open-source software in public law realms.
“It’s incredibly important that people like Rebecca Wexler and Andrea Roth … who are focused on identifying weaknesses in our criminal justice system, are able to find a listening ear on Capitol Hill,” Takano says. “It takes people like them, who are supported by institutions like Berkeley Law, but we’ve also got to have people who are receptive to what they’re doing.”