By Liliana Salgado and Andrew Goudsward
CHANDLER, ARIZONA (Reuters) -A simulation of a dead man created by artificial intelligence addressed his killer in an Arizona court this month, in what appears to be one of the first such instances in a U.S. courtroom.
Made by his family, an AI-generated avatar of Christopher Pelkey spoke in Maricopa County Superior Court on May 1, as a judge prepared to sentence Gabriel Paul Horcasitas for shooting and killing Pelkey in a 2021 road-rage incident.
“It is a shame we encountered each other that day in those circumstances,” the Pelkey avatar says in the video. “In another life, we probably could have been friends.”
The Pelkey avatar appears in the video sporting a long beard and green sweatshirt against a white backdrop. He cautions at the start that he is an AI-version of Pelkey, which is apparent through the gaps in audio and slightly mismatched movement of his mouth.
Pelkey, a U.S. Army veteran, was 37 at the time of the shooting.
The video marked a novel use of AI in the legal system, which has viewed the rapidly growing technology with a mix of fascination and trepidation.
Courts generally have strict rules on the types of information that can be presented in legal proceedings, and several lawyers have been sanctioned after AI systems created fake cases that they cited in legal briefs.
Pelkey’s relatives were given more leeway to present the AI-generated video to the judge at sentencing, given that it was not evidence in the case. Horcasitas, who was sentenced to 10.5 years in state prison, had already been convicted on manslaughter and endangerment charges.
Pelkey’s sister Stacey Wales said she scripted the AI-generated message after struggling to convey years of grief and pain in her own statement. She said she was not ready to forgive Horcasitas, but felt her brother would have a more understanding outlook.
“The goal was to humanize Chris, to reach the judge, and let him know his impact on this world and that he existed,” she told Reuters.
Generative AI, Wales said, is “just another avenue that you can use to reach somebody.”
Wales said she worked with her husband and a family friend, who all work in the tech industry, to create it.
Harry Surden, a law professor at the University of Colorado, said the use of generative AI material in court raises ethical concerns, as others may seek to use those tools to play on the emotions of judges and juries. The content is a simulation of reality, not the verified evidence that courts typically assess, Surden said.
“What we’re seeing is the simulations have gotten so good that it completely bypasses our natural skepticism and goes straight to our emotion,” he said.
(Reporting by Liliana Salgado in Chandler, Arizona and Andrew Goudsward in Washington; editing by Andy Sullivan and Aurora Ellis)