Arizona court orders resentencing over AI-generated victim video

Share
Arizona court orders resentencing over AI-generated victim video

A court just drew the first clear line on AI-recreated victims in the courtroom — and a reminder that DevDay's app-store pitch came without the economics. Two stories this hour.

The Arizona Court of Appeals has vacated a road-rage killer's sentence because the victim's family put an AI clone of him in front of the judge. Gabriel Paul Horcasitas, 55, remains convicted of manslaughter for shooting Christopher Pelkey, 37, at a Chandler red light in 2021, and was sentenced last year to 10½ years — but the term is thrown out and he goes back for resentencing. At the hearing, Pelkey's family rebuilt him from his voice recordings, photos and videos: his sister Stacey Wales supplied the material, her husband assembled the clip, and an AI Pelkey in a grey baseball cap told the man who killed him, "In another life, we probably could have been friends," before offering forgiveness. Trial Judge Todd Lang embraced it at the time — "I loved that AI, thank you for that" — but the appeals court ruled Wednesday that allowing it was fundamental error, writing that "rather than document 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." The distinction is the whole case: a victim-impact statement is evidence of what a real person said, while this was an argument dressed up as testimony, assembled by grieving relatives and evaluated by the judge who imposed the maximum term. The family's motives were sympathetic; the ruling says sympathy is not an evidentiary exception. Expect other states to cite this opinion the first time a family reaches for the same tool — the conviction survived, but AI testimony as a sentencing exhibit just acquired a published appellate rebuke.


OpenAI's DevDay pitch turns ChatGPT into a place to discover and launch software — but not yet a place to get paid. The announcements from the September 29 keynote in San Francisco aim squarely at the app-store model: "Sign in with ChatGPT" launched with 16 partners including Cognition's Devin, Notion and Vercel, the OpenAI Marketplace lined up 32 partners spanning Adobe, Figma and Salesforce, and ChatGPT will now suggest apps mid-conversation and host third-party plugin panels that pass through a human review flow before they ship. The distribution is real — ChatGPT is at 1.2 billion weekly users, up from 1 billion in July, and its Dots agent connects to more than 4,000 apps — alongside GPT-6.1 Sol at a fifth of Astra's token price and a $500-a-month Pro 500 tier. What's missing is the part that makes an app store an economy: OpenAI announced no billing or revenue-sharing system, so developers get discovery without a paycheck. Until that arrives, this is a platform play aimed at Apple and Google, not a marketplace anyone can monetize.

What to watch: how appellate courts beyond Arizona treat AI-recreated victim statements, and whether OpenAI attaches real revenue share to the Marketplace before partners notice the asymmetry.

Should AI-recreated victim statements be barred from court entirely, even when the family asks for it? Tell us in the comments.

Read more

OpenAI drops 372 math results, nearly all from one prompt

OpenAI drops 372 math results, nearly all from one prompt

A dense news day across three fronts: OpenAI published a mountain of mathematics from an unreleased model, SpaceX reportedly wants the debt market to fund its Nvidia orders, and Anthropic restructured its cyber program around three tiers of trust. OpenAI published 372 mathematical result families from an internal frontier model — and says nearly all of them came from a single prompt handed to a single AI agent. The release comprises 722 manuscripts touching hundreds of open questions, plus Lea

Underdog launches a private on-device AI assistant, backed by a16z

Underdog launches a private on-device AI assistant, backed by a16z

The privacy split in consumer AI got a new entrant tonight, GitHub turned code-review benchmarks into a vendor-bias argument, and OpenAI opened its oddest API to everyone. Underdog launched in invite-only beta: an on-device AI assistant that keeps your data on your machine and never charges you a subscription. Self-taught coder and Thiel Fellow Sigil Wen — who moved to Silicon Valley at 17 and lived in an AI hacker house with Andrej Karpathy — built his own inference engine, Husky, to run a 2

Today in AI — October 6, 2026

Today in AI — October 6, 2026

A day of second-order moves: the labs are buying task data from software vendors instead of scraping the web, Waymo is borrowing to fund the robot world, and the biggest bank in the US just put a number on what Anthropic's latest model cost it in risk. Models & Research * OpenAI is training GPT-6 Astra on Ironclad's real contracting work. Ironclad staff helped turn 11 tasks across legal, commercial and procurement work — setting up NDAs, approval workflows, clauses that change by jurisdict

Meta, Walmart and Stripe publish the Personal Agent Protocol

Meta, Walmart and Stripe publish the Personal Agent Protocol

The agent economy is writing its rulebook tonight: one open standard for AI bots at the checkout, a cheaper image model from Google, and Anthropic turning bug-hunting into a tiered product. Meta, Walmart, Stripe and Sierra are publishing an open "personal agent protocol" — a standard that defines how personal AI agents interact with businesses online. The group behind it reads like a cross-section of agentic commerce: Meta, Sierra, Walmart, Stripe, Shopify, Genesys, Rocket, NiCE, Decagon and I