A popular area of discussion in the legal profession right now is how to avoid the improper use of artificial intelligence (AI). When it comes to sentencing, guidance can be found from an Arizona case and a federal judge, and our own good sense as well.

The facts before the Arizona appellate court were unusual: The decedent’s sister, who wanted her brother to have a voice in his assailant’s sentencing (a road rage incident leading to a manslaughter conviction), created a video depicting her brother offering a victim impact statement. Some of the video included real footage of the victim speaking in the past about his faith and spiritual beliefs, but another part of the video was an AI-generated depiction of the victim offering his forgiveness to the defendant. Although there was some effort to be transparent about what was real and what was not, the AI depiction included the victim telling the viewer that this was a “true representation” of him and not how the courtroom had portrayed him.

The judge “loved” the video and felt it was “genuine.” He could see the “obvious forgiveness” on the part of the victim. The judge commended the family that although they had requested a maximum sentence, they allowed the judge to hear from the victim “from his heart as you saw it.” The judge nevertheless imposed the maximum presumptive sentence of 10.5 years.

According to the Arizona Court of Appeals, the “AI video crossed that line” by depicting the victim’s thoughts based on the “imaginings” of the sister. The video “erases the interpretative distance between the family’s belief about what the victim would have said and the victim’s own voice and own opinions.”

Allowing the AI-generated video amounted to fundamental error, and given its clear impact on the sentencing judge, it rendered the sentencing procedure “fundamentally unfair.” State v Horcasitas, ___ P3d ___ (Arizona App Div 1, 9/30/26).

The Horcasitas case analyzes an uncommon scenario, but it also raises the more common question: Are there proper uses of AI at sentencing?

According to a federal district judge in New York, there are benefits to using artificial intelligence that include “reviewing averages, means, and typical sentences in comparable cases.” That said, the judge explained in colorful terms why he could not imagine using artificial intelligence to choose the length of the sentence:

Properly deployed, artificial intelligence can be a useful tool; improperly deployed, artificial intelligence would render judges into useless, foolish puppets—into hapless astronauts named Dave, governed by artificial computers named Hal. This is not a 2001: A Space Odyssey world in which algorithms dictate sentences. This is the real world of flesh and blood and of laws designed to protect our children. We serve justice, not algorithms. [United States v. Stewart, No. 21-CR-42 (WFK), slip op at 6 (E.D.N.Y. Sept. 12, 2025).]

As the judge explained, he imposed what he considered an appropriate sentence because “[t]o ignore the admitted abuse of Defendant’s nephew; to accept the ridiculous assertion Defendant is not a pedophile; to consider the sentence in this case to warrant an average or a mean sentence would result in manifest injustice to the victims of Defendant’s crimes. * * * The behavior of this defendant was not average.” Id.

Sentencing is in many ways a fact-specific process that depends heavily on the views of the sentencing judge. That brings us to our last questionable use of AI: Some defendants may use artificial intelligence to craft their statements in allocution, preferring to offer a more polished statement to the court. AI may be helpful in choosing the precise word to express remorse, contrition or even rehabilitation, all in a style amenable to the judge. But the danger, of course, is that judges are alert to contrived and disingenuous statements. And a defense attorney may be surprised when a client begins with a non-sensical statement. Whether AI-inspired allocution will become popular or strongly discouraged remains to be seen.

We cannot escape the 21st century, but using AI to generate a deceased victim’s impact statement, the appropriate sentence or even statements in allocution may be going too far. (This piece, for better or worse, was written without AI assistance.)

Anne Yantus is a sentencing consultant working with attorneys to promote more favorable sentencing outcomes. Anne credits her knowledge of Michigan sentencing law to the many years she spent handling plea and sentencing appeals with the State Appellate Defender Office. Following her time with SADO, Anne taught a criminal sentencing course at the University of Detroit Mercy School of Law and subsequently continued to write and speak on felony sentencing law while serving as pro bono counsel with Bodman PLC. Anne welcomes your Michigan felony sentencing questions and is happy to arrange a consultation where appropriate. 

Due to the volume of inquiries, Anne is not able to respond to non-attorney requests for assistance or analysis of individual fact situations.

Subscriber Comments

• You must be a subscriber to view comments