An Arizona appeals court has vacated the 10.5-year prison sentence of a man convicted in a fatal road-rage shooting after ruling that an AI-created video of his dead victim made the sentencing hearing fundamentally unfair.
The manslaughter conviction of Gabriel Horcasitas still stands. But the Arizona Court of Appeals ordered a new sentencing after finding the digital recreation of victim Christopher Pelkey was “not reliable” and amounted to a “fundamental error.”
Breitbart News reported that the video, built from an AI recreation of Pelkey’s image and voice, a script written by his sister, real footage, and an aged photo, was used as a victim impact statement and is believed to be the first of its kind used nationwide. The court left the guilty finding intact while sending the punishment back for a do-over.
That outcome lands hard for families who want finality after a killing. It also exposes how quickly a courtroom can drift when novelty outruns the rules meant to keep sentencing grounded in real evidence.
The shooting happened in 2021 in Chandler, Arizona, about 25 miles outside Phoenix. Pelkey, 37, approached Horcasitas’ vehicle during a road-rage confrontation. Horcasitas shot and killed him.
Horcasitas was first convicted in 2023 of manslaughter and endangerment. A new trial followed after a prosecution evidence-disclosure failure. In 2025, a jury found him guilty of manslaughter and endangerment again.
At sentencing in Maricopa County Superior Court, the family presented the AI video. Stacey Wales, Pelkey’s sister, wrote the script. She sought to give her brother a voice in the case over his death.
Judge Todd Lang did not treat the presentation as a cautious experiment. He said, “I loved that AI,” called it “genuine,” and said it gave Pelkey a chance “to speak from his heart.” Then he imposed more than 10 years, longer than the prosecution recommended.
Court rulings that reset high-stakes outcomes are hardly rare in Arizona or at the national level, as when an Arizona voter citizenship case headed toward Supreme Court review after a circuit fight over election rules. The difference here is the tool that pulled the sentence apart: a machine-made stand-in for a man who could not speak.
The Arizona Court of Appeals acted on a Wednesday, vacating the 10.5-year term and ordering resentencing while upholding the manslaughter conviction. No resentencing date has been set.
The New York Post reported that the panel found the AI video “not reliable” and a “fundamental error” that prejudiced the judge and left the sentencing “fundamentally unfair.”
No prior Arizona case had addressed the admissibility of an AI-created depiction of a victim offered as a victim impact statement. The court looked to Arizona Supreme Court guidance in State v. Rose, the death-sentence case of Edward James Rose for the 2007 killing of a Phoenix police officer. In Rose, photographs of the officer’s sons at the grave came “uncomfortably close” to the line but did not cross it.
The AI presentation did.
“By contrast, the AI video crossed that line. Unlike the photographs in Rose, which captured a particularly poignant moment for the young surviving victims, the AI video here (excluding the embedded ‘real’ videos of the victim) does not reflect actual events,”
the court wrote.
It went further on the source of the words and images.
“Indeed, rather than documenting an event or recording a particular moment, the AI video presents a depiction of the victim and his thoughts created from the imaginings of the victim’s sister.”
That is the core problem in plain terms. Photographs of a real funeral moment are one thing. A scripted AI avatar speaking as if it were the dead man is another. One records what happened. The other invents what someone wishes had been said.
The record does not leave much doubt that the video mattered. Reuters reported the appeals court’s blunt assessment of prejudice and the role the clip played at sentencing.
“While the record does not reflect precisely how the AI video factored into the sentencing calculus, there is no question it played a role.”
That line came from Judge D. Steven Williams. The panel’s conclusion was direct.
“The trial judge’s consideration of the AI video so prejudiced Horcasitas as to render the sentencing procedure fundamentally unfair.”
Judge Lang’s own words at the original hearing made the risk obvious. He did not merely allow the video. He praised it. He treated a computer recreation as heartfelt testimony from the grave. Then he handed down a longer term than prosecutors asked for.
Victims’ families have every right to be heard. Courts have long allowed impact statements for that reason. But impact statements are supposed to come from living people describing real loss, not from software puppeteering a likeness of the dead.
When a judge calls an AI fabrication “genuine,” the system stops looking like sober judgment and starts looking like a tech demo with a prison term attached. Other high-profile fights over whether convictions and sentences hold up under hard review, including when a federal judge dismantled an attempt to overturn a major conviction, turn on the same basic demand: the process has to be clean enough to trust.
The package was not a simple photo or a short clip of Pelkey while he was alive. It mixed real video of the victim with an AI recreation of his image and voice. Wales wrote the script. An old-age filter photo was part of the presentation.
Reporting on the case noted the video showed Pelkey appearing to address forgiveness toward his killer. That emotional punch is exactly why the format is so dangerous in a courtroom. Jurors and judges are human. A lifelike digital ghost can move them in ways a sister’s live statement, powerful as it is, does not pretend to be the victim himself.
The appeals court drew the line at reliability. The AI portion did not document an event. It depicted thoughts “created from the imaginings” of a grieving relative. Grief is real. The manufactured monologue is not evidence of what Pelkey believed, intended, or would have said.
Law-and-order conservatives should not shrug at that distinction. Accountability for a killing requires a process the public can defend. If sentencing rests on synthetic speech, every future case becomes a contest of who can hire the better deepfake team.
Arizona has seen sharp federal rulings on enforcement and sovereignty before, including when a federal judge cleared border wall construction on Arizona tribal land. Different subject. Same underlying need: institutions that follow clear rules instead of improvising under pressure.
Horcasitas remains convicted of manslaughter and endangerment. He is not walking free on a technicality that erases guilt. The jury’s finding survived. Only the sentence was thrown out.
Still, the practical result is delay. A killer who already went through a 2023 conviction, a new trial after a disclosure failure, and a 2025 reconviction now gets another sentencing hearing. Pelkey’s family must watch the punishment phase reopen because the trial court welcomed an untested AI spectacle and then celebrated it on the record.
That is the cost of treating courtroom procedure like a sandbox for new tools. Prosecutors failed to disclose evidence once and forced a retrial. The sentencing judge then embraced an AI video with no Arizona precedent. The appeals court had to clean up the mess while leaving the conviction in place.
Emergency and appellate fights over process, and whether prior orders stand, are familiar terrain, much like when the Supreme Court left Missouri’s old congressional map in place after rejecting a rushed challenge. Procedure is not a sideshow. It is how a free society keeps punishment legitimate.
Victim impact statements exist so judges hear the human cost of crime. Sisters, parents, and spouses should be able to stand up and describe what was stolen from them. Nothing in this ruling cancels that right.
What it blocks is the leap from testimony to theater. An AI avatar speaking in a dead man’s voice, reading lines written after his death, is not a victim “speaking from his heart.” It is a production. Calling it genuine does not make it so.
State v. Rose already warned how close emotional exhibits can come to improper influence. Photographs of children at a grave were uncomfortable but still tied to a real moment. The Horcasitas video left that ground. It offered imagined thoughts as if they were the victim’s own.
Courts that ignore that difference invite chaos. Defense teams will demand equal access to synthetic mitigation videos. Prosecutors will answer with AI recreations of crime scenes and dying words never recorded. Judges who “love” the tech will struggle to explain why one deepfake is moving and another is barred.
The Arizona Court of Appeals refused to pretend the problem was abstract. It said the video prejudiced the defendant, was not reliable, and made the sentencing fundamentally unfair. The conviction held. The sentence did not.
Horcasitas will be resentenced under rules that treat evidence as evidence, not as a scripted digital performance. Pelkey’s family can still be heard in person, without a machine wearing his face.
Justice for a road-rage killing should rest on facts, law, and real human testimony, not on a courtroom that mistakes an algorithm’s mask for the voice of the dead.
By signing up, you agree to receive newsletters and promotional content from American Frontline News and selected publications in the American Digest Media Network, operated with Patriot Mom Digest LLC, and you accept our Terms of Use and Privacy Policy. You may unsubscribe at any time.
By signing up, you agree to receive newsletters and promotional content from American Frontline News and selected publications in the American Digest Media Network, operated with Patriot Mom Digest LLC, and you accept our Terms of Use and Privacy Policy. You may unsubscribe at any time.