WASHINGTON (CN) — As law enforcement agencies turn to AI-powered tools to streamline policing and evidence gathering, lawmakers should step in to help manage the potential drawbacks of such fast-developing technology, a panel of experts said Wednesday.
Artificial intelligence is quickly pushing its way into police organizations across the country via tools such as facial recognition and gunshot detection programs. AI can also help augment existing human-led law enforcement activities like social media monitoring.
Despite the recent discourse surrounding AI and algorithmic models, some of these programs have already been in use for years. According to the Government Accountability Office, an independent government watchdog, federal agencies such as the FBI and U.S. Customs and Border Protection have been using facial recognition services since at least 2018.
Proponents of such technology say that the increased preponderance of artificial intelligence will make law enforcement more effective and efficient.
Miami Police Department Assistant Chief Armando Aguilar told the Senate Judiciary Committee during a Wednesday hearing that artificial intelligence helped Miami police execute one of “the greatest turnaround stories in law enforcement.” The city’s homicide rate has fallen dramatically over the last decade, he said, a feat helped out by AI.
Aguilar, who also serves on the White House’s National AI Advisory Committee’s law enforcement subgroup, explained that Miami detectives who use artificial intelligence tools are two-thirds more likely to solve violent crimes.
Miami has also been transparent about its use of AI tools, such as facial recognition, Aguilar said. Police developed a facial recognition policy in collaboration with the community and kept the public up to date with successful arrests made using the technology.
“The policy that resulted from our efforts created a narrow framework within which we would use [facial recognition],” Aguilar told lawmakers.
Miami police established safeguards on facial recognition technology, emphasizing that facial matches do not represent probable cause for arrest — treating them instead like an anonymous tip that must be corroborated with physical or circumstantial evidence. The Miami Police Department also does not use facial recognition in real time and does not “substantively manipulate” photographs taken using facial recognition technology, Aguilar said.
While facial recognition and other algorithm-based law enforcement tools can speed up the process of police work, their implementation leaves a lot to be desired, said Karen Howard, director of the Government Accountability Office’s science, technology assessment and analytics division.
Some of the potential pitfalls are rooted in human error, she said, such as poor-quality crime scene evidence being ingested by AI tools, or improper use by human analysts. Other issues lie in the data on which AI models are trained, which can be “unknown or potentially biased,” Howard added.
A lack of public trust in artificial intelligence could also impact its use, she said.
Meanwhile, Rebecca Wexler, a law professor at the University of California at Berkeley, lamented what she said were “troubling obstacles” to scrutinizing the use of AI tools in criminal legal proceedings.
Wexler criticized some vendors of AI tools which she said can freely refuse access to researchers hoping to conduct independent reviews of their services, something she said averts scrutiny of potential flaws.
AI vendors are also able to shield their tools from review by citing trade secret privilege, Wexler told the committee, by refusing to share the details of their technology with courts, witnesses and criminal defendants.
“This should not be happening,” she said. “We need criminal defense counsel and experts, as well as judges and prosecutors, to scrutinize AI systems to ensure accuracy and fairness.”
Such cross-examination is particularly important given the potential hazards of allowing evidence gathered by AI tools to be used in law enforcement unchecked, Wexler argued, pointing out that erroneous facial recognition and gunshot alerts have led to wrongful arrests and imprisonment — and that using AI in capital cases could even lead to wrongful death.
“We need fair and open proceedings to expose these kinds of costly mistakes and discrepancies,” she said.
Wexler said that Congress should require AI vendors to allow independent audits of their tools, and that lawmakers should restrict federal law enforcement agencies from buying such services unless they are made available to such review.
Lawmakers should also clarify the law to ensure that companies can’t cite trade secret privileges in federal criminal cases if required to disclose such information under court order, she added.
Howard said that Congress should increase transparency on the testing performance and use of algorithms used in police work. Such a change would improve law enforcement understanding of AI training data and could help police identify the best performing algorithms.
Lawmakers should also facilitate “the adoption of standards for appropriate use” of AI tools for law enforcement, she added, helping to reduce the use of low quality evidence and hike public confidence in the technology.
Senators in attendance Wednesday appeared to be of two minds about the use of AI in law enforcement.
Senator Cory Booker said the expanded use of artificial intelligence “raises a lot of questions” but also “raised a lot of excitement, for me, about the possibilities.”
AI poses “potential problems when it comes to our criminal justice system,” said the New Jersey Democrat, “where often a person’s liberty and a person’s life are at stake.”
Missouri Senator Josh Hawley, meanwhile, sought to tamp down some fears about the power of artificial intelligence, arguing it serves as a tool for human law enforcement and is not an autonomous decision-making entity.
AI, said Hawley, a Republican, provides “impressive time saving, crime solving and justice serving tools,” pointing to its ability to quickly narrow down matches on facial recognition technology.
“If the early stages of a criminal investigation is like looking for a needle in a haystack, then artificial intelligence can provide a magnet for some of our criminal investigators in police forces around the country,” Hawley said. “It’s not Robocop. It’s not the Terminator or the Matrix, it’s not Ultron and it’s not even WALL-E.”
Congress has explored avenues for regulating artificial intelligence for months in an effort to get its arms around the technology’s lightning pace. The White House, similarly, has unveiled its own set of standards for managing AI safety and security.
Follow @BenjaminSWeiss
Subscribe to Closing Arguments
Sign up for new weekly newsletter Closing Arguments to get the latest about ongoing trials, major litigation and hot cases and rulings in courthouses around the U.S. and the world.
Additional Reads
-
Sudan war marks one-year anniversary April 15, 2024 -
California certifies March 5 primary election results April 12, 2024
