Arizona appeals court vacates a sentence that relied on an AI victim video

An Arizona appeals court has thrown out a prison sentence because the judge relied on an AI video of the person who was killed. The opinion in State v. Horcasitas was filed on 30 September 2026. NBC News reported the decision on 4 October, and CBS, with the Associated Press, on 5 October. This is that ruling, read from the opinion.

The case is No. 1 CA-CR 25-0191, in the Arizona Court of Appeals, Division One. Presiding Judge D. Steven Williams wrote the opinion. Judge Anni Hill Foster and Vice Chief Judge David D. Weinzweig joined. The appeal is from the Superior Court in Maricopa County, the Honorable Todd F. Lang. The defendant is Gabriel Paul Horcasitas. The opinion does not name the victim or the victim’s sister and brother. It refers to them by those roles.

The header reads “AFFIRMED IN PART; VACATED AND REMANDED IN PART.” Paragraph 54 says: “we affirm the manslaughter conviction but vacate the sentence for manslaughter, and remand for resentencing.”

What the court decided

The opinion describes a fatal road-rage case and a manslaughter conviction after the jury rejected self-defense. Paragraph 1 says the sentencing court relied on an AI video it called “genuine” and “speaki[ng] from his heart,” and that the appeals court vacates the manslaughter sentence “because the AI video was not reliable.” The conviction stands. A separate challenge to excluded text messages from the victim’s phone fails.

The superior court imposed a presumptive 10.5 years for manslaughter and a concurrent presumptive 2.25 years for endangerment. The conclusion vacates the manslaughter sentence. It does not say the endangerment term was vacated or affirmed. CBS called the term 10 years. The opinion’s figure is 10.5.

There was no objection at sentencing, so review is for fundamental error. Victim impact evidence is allowed “unless it is so unduly prejudicial that it renders” sentencing “fundamentally unfair.”

Why this was not an ordinary statement

The Victims’ Bill of Rights lets a victim be heard at sentencing. The opinion cites Arizona Constitution article 2, section 2.1(A)(4), and A.R.S. section 13-4426. Speakers in that role are not witnesses: no oath, no cross-examination, and no advance disclosure, under section 13-4426.01. The right “is not unlimited.” Sentencing may use reliable relevant evidence, including hearsay, but due process still requires a “minimal indicia of reliability” and a sentence “on the basis of accurate information.” Paragraph 48 says: “No Arizona case has addressed the admissibility of an AI-generated depiction of a victim offered as victim impact evidence.”

Family members spoke. Several asked for the maximum sentence, which, because the State alleged no aggravating factors, was the presumptive term of 10.5 years. The brother read words the victim had written. The sister, the opinion says, told the court she was imagining what he would have said. The AI video then played. It began as a “version” of the victim “recreated through AI that use[d] [his] picture and [his] voice profile,” with a real video embedded, then said it “is a true representation of who [he] was . . . not how the courtroom portrayed [him].” It addressed Horcasitas in part: “I believe in forgiveness and in God who forgives.” The judge said, “I loved that AI,” that it “was genuine,” and that the family “allowed [victim’s first name] to speak from his heart as you saw it.” He imposed 10.5 years: “That’s more than the State asked for.” The prosecutor had asked for 9.

The holding, in the opinion’s words

Paragraph 50 says “the AI video crossed that line.” The AI video, “excluding the embedded ‘real’ videos of the victim,” “does not reflect actual events.” “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.”

Paragraph 51 says the video “presents the statement … as coming directly from the victim. It erases the interpretive distance between the family’s belief about what the victim would have said and the victim’s own voice and opinions.” The court’s next sentence is: “No disclaimer could cure that error.”

Paragraph 53 says the embedded real footage was permissible, and that “allowing the presentation of the AI video constituted fundamental error” and “so prejudiced Horcasitas as to render the sentencing procedure fundamentally unfair.” The court vacates the manslaughter sentence and remands for resentencing.

CBS, with the Associated Press, wrote: “In what’s believed to be a first in U.S. courts.” That is the outlet’s characterization. The opinion itself does not say the ruling is a national first.