Skip to main content

Arizona Court Tosses 10.5-Year Sentence Over AI-Generated Victim Impact Video

by Alexandra Agraz | Oct 02, 2026
Judge’s gavel rests on a wooden sound block in a courtroom scene, with a U.S. flag and clock in the background.
Photo Source: Adobe Stock Image

An Arizona appeals court has thrown out a 10.5-year manslaughter sentence after finding that a judge improperly relied on an AI-generated victim impact video that recreated shooting victim Christopher Pelkey speaking at the defendant’s sentencing.

In a unanimous published opinion issued September 30 in State v. Horcasitas, the Arizona Court of Appeals, Division One, vacated Gabriel Paul Horcasitas’ sentence and ordered a new sentencing hearing while leaving his manslaughter conviction intact. The panel found that the artificial intelligence recreation of Pelkey lacked the reliability required for information used to determine a criminal sentence.

The appeal followed Horcasitas’ second trial in Pelkey’s fatal 2021 road-rage shooting. A jury rejected his claim that he acted in self-defense and convicted him of manslaughter. Horcasitas challenged both the exclusion of text messages recovered from Pelkey’s phone and the use of the AI-generated victim presentation at sentencing. The appeals court rejected the evidentiary challenge but agreed that consideration of the synthetic video violated his right to a fair sentencing proceeding.

At sentencing, Pelkey’s family presented a video that used AI to recreate his appearance and voice. The synthetic version addressed his family, the court, and Horcasitas while also incorporating authentic video footage of Pelkey.

The recreation identified itself as a version of Pelkey created using his photograph and voice profile. It later addressed Horcasitas and expressed forgiveness, reflecting beliefs that Pelkey’s sister said she thought her brother would have expressed if he had been able to speak at the hearing.

Maricopa County Superior Court Judge Todd Lang reacted favorably to the presentation. According to the appellate opinion, Lang said he “loved that AI,” described the recreation as “genuine” and referred to Pelkey’s apparent forgiveness of Horcasitas as coming from his “heart.”

Pelkey’s family asked for the maximum sentence available, while prosecutors sought a slightly mitigated nine-year term and the defense requested seven years. Because prosecutors had not alleged aggravating factors, the presumptive 10.5-year sentence was also the maximum available for the manslaughter conviction.

Lang imposed the 10.5-year term and acknowledged that it was more than prosecutors had requested.

Arizona’s Victims’ Bill of Rights gives crime victims broad protections in the criminal justice system, including the right to be heard before sentencing. Victim impact statements are also treated differently from ordinary trial testimony. Victims do not testify under oath or face cross-examination, and sentencing judges have wide discretion to consider information that would not necessarily be admissible before a jury.

But sentencing information still must have at least a minimal level of reliability because defendants have a due process right to be sentenced using accurate information. The appeals court found that the synthetic portions of the video failed that standard.

The panel emphasized that those portions did not document anything Pelkey actually said or did. Instead, they presented his sister’s conception of what he would have thought and said through Pelkey’s recreated face and voice, eliminating the distance between the family’s understanding of his views and Pelkey’s actual views.

The State argued that the recreation was sufficiently reliable because Pelkey’s sister knew him closely and the sentiments attributed to him were consistent with his writings, beliefs and values. The appellate panel was not persuaded.

Although the video disclosed that artificial intelligence had been used, the court found that disclosure did not make the synthetic statements reliable. The recreated Pelkey subsequently described himself as a true representation of who Pelkey was, while the video supplied facial expressions, a voice, and statements that Pelkey had never recorded. The panel concluded that the disclaimer did not cure the resulting reliability problem.

For guidance, the judges looked to State v. Rose, an Arizona Supreme Court decision involving emotionally powerful victim impact evidence. The court in Rose allowed photographs depicting the grief of surviving family members despite finding that the presentation came close to the boundary between permissible victim impact evidence and material that could make sentencing fundamentally unfair.

Unlike the photographs in Rose, which documented events that actually occurred, the synthetic portions of Pelkey’s video created speech and conduct that never occurred. The AI presentation attributed those words and expressions to a person who could no longer confirm what he believed or would have said.

The appellate court also found that the video played a role in Lang’s sentencing decision, pointing to his comments immediately after the presentation. Although the record did not show how much weight he ultimately gave the AI recreation, the panel found that he relied on it in imposing the sentence.

Because Horcasitas had not objected to the AI video at sentencing, the appeals court reviewed the issue for fundamental error rather than under the standard applied to a preserved objection. The panel still found the error serious enough to render the sentencing proceeding fundamentally unfair.

The ruling applies specifically to the synthetic portions of the victim impact presentation. The panel distinguished those sections from genuine footage of Pelkey included in the same video and focused on new statements, facial expressions and conduct created by AI and attributed directly to a victim who could no longer confirm them.

Other courtroom uses of artificial intelligence remain unresolved in Arizona, including technology used to restore authentic recordings, present a person’s actual words or create demonstrative evidence. The court described State v. Horcasitas as the first Arizona case to address an AI-generated depiction of a victim offered as victim impact evidence.

The case had already returned to trial once before the latest appeal. Horcasitas’ first manslaughter conviction was set aside after the trial court found that prosecutors failed to timely disclose evidence recovered from Pelkey’s phone that included material supporting his self-defense arguments. The Court of Appeals upheld the new-trial order before a second jury again convicted him.

On the separate text-message challenge in the latest appeal, the panel upheld the trial court’s evidentiary rulings, including the exclusion of a message concerning Pelkey’s mental state. That ruling left Horcasitas’ second manslaughter conviction intact.

With the conviction affirmed and the sentence vacated, Horcasitas will return to Maricopa County Superior Court for resentencing.

Share This Article

If you found this article insightful, consider sharing it with your network.

Alexandra Agraz
Alexandra Agraz is a former Diplomatic Aide with firsthand experience in facilitating high-level international events, including the signing of critical economic and political agreements between the United States and Mexico. She holds dual associate degrees in Humanities, Social and Political Sciences, and Film, blending a diverse academic background in diplomacy, culture, and storytelling. This unique combination enables her to provide nuanced perspectives on global relations and cultural narratives.

Related Articles

Judge’s gavel held above its sound block on a desk, symbolizing the federal court sanctions against four attorneys.

Federal Judge Removes All Four Lawyers From Case Over Fake AI Citations

A federal judge in Mississippi removed all four attorneys from a civil lawsuit, imposed $8,000 in fines, and barred two lawyers from appearing in the district for two years after both sides submitted court filings containing fake legal citations generated by artificial intelligence.U.S. District Judge Sharion Aycock issued the sanctions... Read More about Federal Judge Removes All Four Lawyers From Case Over Fake AI Citations

A wooden judge’s gavel rests on its sound block in a courtroom, with a blurred person writing documents in the background.

Man Gets 18 Months in First U.S. AI Music Streaming Fraud Case After $8M Royalty Scheme

A federal judge has sentenced a North Carolina musician to 18 months in prison for using AI-generated songs and thousands of automated accounts to produce billions of fake streams and collect more than $8 million in royalties, bringing the first U.S. criminal case over AI-assisted music streaming fraud to sentencing.U.S.... Read More about Man Gets 18 Months in First U.S. AI Music Streaming Fraud Case After $8M Royalty Scheme

Hands typing on a laptop beneath a glowing AI prompt interface that displays “Prompt:” and “Generate.”

Federal Appeals Court Blocks Charge Over Private Possession of AI-Generated Child Sexual Abuse Images

A federal appeals court has upheld the dismissal of a charge against Wisconsin resident Steven Anderegg for privately possessing obscene AI-generated sexual images depicting fictional children in his home, finding that existing Supreme Court precedent protects that narrow form of possession under the First Amendment. Separate federal charges accusing him... Read More about Federal Appeals Court Blocks Charge Over Private Possession of AI-Generated Child Sexual Abuse Images

The U.S. Court of Appeals for the Tenth Circuit building in Denver, Colorado.

Federal Appeals Court Moves to Crack Down on AI Filings After Lawyers Cite Fake Cases

A federal appeals court is moving to tighten its rules on artificial intelligence after judges encountered court filings containing fabricated cases, false quotations and other AI-generated legal errors.The U.S. Court of Appeals for the Tenth Circuit proposed a new rule September 18 that would require lawyers and self-represented litigants to... Read More about Federal Appeals Court Moves to Crack Down on AI Filings After Lawyers Cite Fake Cases

Hands type on a laptop displaying the ChatGPT interface.

New Mexico Lawyer Fined $5,000 After Using ChatGPT to Draft Murder Appeal With Fake Witnesses

A New Mexico defense lawyer has been fined $5,000 after using ChatGPT to help draft a murder appeal, failing to check what it produced and filing a brief that included four witnesses who never existed, false testimony, and misrepresented legal authority.The New Mexico Supreme Court held Santa Fe attorney Stephen... Read More about New Mexico Lawyer Fined $5,000 After Using ChatGPT to Draft Murder Appeal With Fake Witnesses

Legal term

Search articles and Legalpedia

Subscribe to Newsletter