BOULDER — In modern society, we are surrounded by video. In criminal cases, video evidence is often important in court.
However, a local study believes the US criminal justice system is ill-prepared for the rise of artificial intelligence (AI) and how it may affect video evidence.
The first report from the newly established Visual Evidence Lab at the University of Colorado Boulder (CU) stems from a discussion earlier this year in which experts analyzed the challenges and solutions related to AI in the courtroom.
The report begins by examining the scope of the problem, citing statistics from the U.S. Department of Justice's Office of Legal Assistance, which estimates that approximately 80% of criminal cases in 2016 included video as evidence. However, the report's authors said there is no uniform guidance in the country when it comes to such extensive video evidence.
“Video can be important evidence in civil and criminal trials, but without consistent guidelines on how to use that material, we are at risk of unfair verdicts,” said Sandra Ristowska, founding director of the Visual Evidence Lab and associate professor of media studies at the University of Boulder. “There is no one easy, perfect solution, and problems need to be addressed at multiple layers.”
The report goes on to argue that “the legal system's unregulated approach to video evidence risks giving rise to contradictory and even erroneous interpretations.”
The study highlights three main concerns regarding AI and the criminal justice system. These include problems with detecting and verifying AI-generated media, uncertainty about what technological enhancements will be allowed in court, and the potential for widespread “deepfake defenses” in which real footage can be challenged in court as fake.
“Currently, in a courtroom, jurors may give little or no weight to real footage that could be important evidence,” Ristowska said. “We need to have a common understanding of truth and facts in order to continue to live in a democratic society and really hold on to this important idea of equal and fair justice.”
Ristowska worries that the rise of AI will lead to even more doubts about the authenticity and integrity of all images, even real ones.
“That is why consistent guidelines are paramount for courts to recognize and uphold civil and human rights in the age of AI,” Ristowska said.
Ultimately, this was key to the recommended solutions developed in the report, which identified four priorities. These include developing infrastructure to store and access video evidence, training judges, providing science-based jury instructions, and implementing safeguards when AI-based evidence is admitted in court.
national news
AI safety report warns industry is 'structurally unprepared' for increased risks
Chief Deputy District Attorney Joel Zinke works for Arapahoe County and agreed with the need for some regulation related to AI and the criminal justice system.
“Frankly, we are past the point where we should be doing this, but we need to put policies in place around the use of AI,” Zink said. “I think that's going to require more cooperation and partnerships and bridge-building with state legislatures, the federal government and local law enforcement.”
Zink said video evidence is one of the best ways for jurors to decide on a case, and investigators have to sort through hours of video in a single case. In this way, he believes AI can help.
“Any kind of tools that allow the prosecutor, or me if I'm working on a case, to more quickly identify key points in a video that's been going on for hours and hours, we need to look at leveraging that,” Zink said. “On the other hand, unfortunately, we have witnessed actual incidents, and I cannot speak to actual incidents, but we have witnessed situations where child predators create what appears to be very realistic child sexual abuse material. They are generating this stuff based on images of the faces of actual children in our community.”
Within the 18th Judicial District, Zink has never seen an AI-altered video submitted as evidence in court. He is concerned about the “deepfake defenses” described in the report from the University of Boulder.
“While we should all be concerned that deepfakes are happening and that they can be misused, it is equally concerning to me as a prosecutor that juries will no longer accept legitimate video evidence for what it is,” Zinke said. “This has been some of the strongest evidence in our case for a very long time, and we want to continue to rely on it.”
Zinke believes that AI has the potential to be extremely helpful within the criminal justice system, but he has seen examples across the country where it can be harmful.
“This incident is occurring in many places across the country and occurs with enough frequency that it should concern us all. As lawyers, people are using large language models to try to write court motions and briefs, but then submit them to court without properly checking all the citations for references to a particular incident,” Zink explained. “The court tries to check the citation, and in the end, it turns out it's not the actual case. Likewise, the citation doesn't exist…It's a breach of your duty as an attorney to file a motion with the court without actually checking the citation.”
Zink believes AI is something the criminal justice system will have to adapt to, both for good and for bad.
“I really think we should be cautious and we should have rules and regulations, and we shouldn't be afraid of this,” Zink said. “If we can find a way to leverage this as a society, I think there are incredible possibilities ahead of us.”
Denver7 | Your Voice: Contact Colette Bordelon
Denver7's Colette Bordelon covers stories that affect all Colorado communities, but specializes in reporting on issues impacting crime, justice, climate and the environment. If you would like to contact Colette, please send an email by filling out the form below.
