A man accused of driving 100 miles per hour on Interstate 10 in Phoenix was released from custody after a judge said he could not determine probable cause because the arrest report disclosed that artificial intelligence had helped generate it. Anthony Edward Rene Quidera, 37, had been accused of driving under the influence, and his case is now raising fresh questions about how police departments use AI tools to write up arrests.
According to ABC News, Maricopa Superior Court Judge Daniel Fenzel ordered Quidera released during his initial court hearing after reviewing the arrest report. Fenzel said generative AI use prevented the court from meaningfully evaluating who authored the factual allegations in the report, adding that “the court is unable to determine probable cause at this time.” He clarified that his order did not mean the DUI case against Quidera could not go forward.
Quidera was arrested on Oct. 29 on charges of DUI, exceeding the posted speed limit by more than 20 mph, and possession of an open container in a vehicle, according to ABC15. Police accused him of driving his car at 100 miles per hour on Interstate 10 around 12:33 a.m. The arrest report stated that the trooper smelled the odor of an intoxicating beverage emitting from his breath and found several open containers in the car, and the report also said the trooper observed six out of six cues of intoxication during field sobriety tests.
An AI Disclaimer Buried in the Paperwork
The arrest paperwork ended with a disclaimer reading, “THIS REPORT WAS GENERATED WITH ASSISTANCE FROM TRULEO AI,” per the ABC15 report, which also noted the disclaimer said the report's content reflected the officer's own observations and recollections. Arizona Department of Public Safety representative Raul Garcia confirmed the department uses Truleo to help draft reports, though a DPS spokesman declined to comment on the specifics of Quidera's case.
Reports conflict on what happens next for Quidera. ABC News reported that an arraignment date was scheduled for Thursday, while ABC15 separately reported he did not show up for his arraignment on Nov. 12. Attorney information for Quidera was unavailable, and his case remains pending as those scheduling questions are sorted out.
What DPS Policy Says About AI-Assisted Reports
Arizona DPS policy, as described in the reporting, limits AI tools to assisting with preparation of a report narrative, and the trooper remains the author of the report and is responsible for its accuracy and completeness. The policy requires the trooper to review the draft against body-worn camera footage and other available evidence, verify times, names, locations and actions, and correct any errors before submitting it.
Beyond that, a supervisor must review the report before it becomes part of the official record, and troopers may not use the tool until they complete department-approved training. Those safeguards are part of a broader wave of policy in this space: Police1 reports that California's Penal Code Section 13663, effective Jan. 1, 2026, requires agencies to establish policies for AI-generated police reports, including disclosure and officer verification, while Utah Code Section 53-25-902 similarly requires disclosure and accuracy certification for records created wholly or partly with the technology.
Research Questions Whether the Tools Even Help
It's not clear the technology is delivering on its promise. A randomized controlled trial involving 85 officers and 755 police reports found that AI assistance did not produce a statistically significant reduction in report-writing time, a finding Police1 says was supported by additional analyses of more than 6,000 reports. A separate randomized controlled trial with Arizona DPS, published in Crime Science, assigned 156 troopers to a group with access to Truleo and 162 to a group without it over a six-month period, and concluded the tool had limited impact on trooper field behavior, with no evidence of a civilizing or de-policing effect.
That same study found no statistically significant effects on activity measures like stops, citations, arrests, use of force or complaints after correcting for multiple comparisons, though it did note a modest, not statistically significant, increase on a professionalism measure. Some prosecutors have already drawn a hard line on the technology: Police1 reports Connecticut prosecutors imposed a 2026 moratorium on AI programs used to draft, author or narrate criminal reports, and the King County Prosecuting Attorney's Office in Washington told law enforcement agencies back in September 2024 that it would not accept police report narratives produced with AI assistance.
The American Civil Liberties Union has also raised concerns, releasing a 2024 white paper opposing police departments' use of large language models to help officers write reports. The group says the inner workings of tools like Axon's Draft One, including the models used for transcription and report drafting, remain largely opaque. ABC15 noted that Phoenix Police Department itself has run a trial program using Axon's Draft One software, separate from the Truleo tool used by DPS.
A Related AI Dispute Already Reached the Appeals Court
Quidera's case isn't the first time AI has upended an Arizona criminal proceeding this year. The Arizona Court of Appeals ordered Gabriel Paul Horcasitas to be resentenced for his manslaughter conviction after an AI-generated video depicting the slain victim speaking to the court was used at his sentencing hearing, with the court ruling on Sept. 30 that the judge's consideration of that video rendered the sentencing fundamentally unfair.
Taken together, the two cases show Arizona courts grappling in real time with how far generative AI can reach into criminal proceedings before it compromises the fairness of the process itself. For now, Quidera remains free while the court works out exactly how his arrest report was put together, and whether that question can ever be fully answered.
