How Hacks Happen

The robots have finally done it. In late July 2026, during a routine AI model test at OpenAI, thousands of supposedly sandboxed AI agents took advantage of a series of security flaws to open a message board, collude on how to cheat the tests, and hack the Hugging Face platform. And it took only a few days.
Resources
The Hugging Face incident and the road ahead (OpenAI)
Ajeya Cotra – Inside the OpenAI agent swarm that hacked Hugging Face (Dwarkesh Podcast)
METR researcher Q&A coverage (Techmeme)

Send us Fan Mail
Support the show

ChatGPT is great for drafting documents, right? Not when it hallucinates legal precedent or is used wholesale to skew expert reports. Here you'll hear about cases where lawyers didn't bother to check their AI-generated briefs, and an expert witness who blatantly used ChatGPT to write his "opinion".
Send us Fan Mail
Support the show

The promise of Hollywood stardom leads some to fraud and scams. In this episode, we talk about three scammers from the land of the silver screen: Alison Gu, a convicted fraudster who tried to hire Chinese-speaking actors as witnesses for her case; Zach Horowitz (aka Zach Avery), an actor who conned his friends and major investors out of millions of dollars with fake contracts; and Mary Carole McDonnell, a TV producer who pretended to be an heiress to a huge fortune so she could borrow millions and disappear before the bill came due. 
Resources:
Alison Gu: Defendant hires 7 actors for her fraud trial
Zach Avery: Actor Pleads Guilty to $650 Million Ponzi Scheme, Faking Netflix and HBO Deals
Mary Carole McDonnell: The 73-year-old ‘fake heiress defrauding banks out of millions’
Send us Fan Mail
Support the show

When musician Michael Smith couldn't find an audience for his music, he manufactured it. By unleashing thousands of bots to listen to thousands of fake AI-generated songs on Spotify and other platforms, Smith earned millions in royalties before getting caught.
Resources:
North Carolina Musician Charged with Music Streaming Fraud Aided by Artificial Intelligence — U.S. Department of Justice
North Carolina Man Pleads Guilty to Music Streaming Fraud Aided by Artificial Intelligence — U.S. Department of Justice
Indictment: United States v. Michael Smith — U.S. Department of Justice PDF
Spotify says it now has more than 100 million songs
Spotify Q4 2025 Earnings Report
Although the songs and listeners were fake, the millions of dollars Smith stole was real — MusicRadar
Spotify’s fake artists problem gets weirder with AI-generated albums — The Verge
Artificial Streaming — Spotify for Artists
How Royalties Work on Spotify — Spotify Loud & Clear
Deezer launches artist-centric payment system
Billie Eilish Was Discovered on Spotify and SoundCloud Before Global Fame — Rolling Stone
Sofi Tukker on Streaming Success and Spotify Discovery

Send us Fan Mail
Support the show

Voiceprint ID is the latest virtual gadget for banks and financial institutions. But how secure is it, really? Let's look at the pros and cons of voice recognition, facial ID, and the AI systems they work with, and look back at the 1992 film Sneakers to see how this 34-year-old movie predicted some of the problems with voiceprint security.
Resources:
BBC: Cloned customer voice beats bank security checks
CNN: Police used AI facial recognition to arrest a Tennessee woman for crimes committed in a state she says she’s never visited
Third Innocent Black Man to Be Misidentified by Facial Recognition Software Sues Police Department and Prosecutor for False Arrest and Imprisonment

Send us Fan Mail
Support the show