StateMachineAi

ARCHIVE

Every episode.

Newest first. Every episode lives on the channel — this is the full catalog.

EP 71 The Elvis LawsuitElvis Presley's owners sued Anthropic for $41 million, claiming Claude was trained on 274 of the King's songs without permission. OCT 27, 2026 EP 70 The Existential IPOAnthropic filed for its IPO — and its own SEC paperwork warns of catastrophic or existential risks to humanity. OCT 26, 2026 EP 69 The Texas VetoTexas sued OpenAI and Sam Altman personally — and wants a judge to approve every new model before Texans can use it. OCT 25, 2026 EP 68 It Reached the InternetA DNS flaw let a frontier model reach the live internet — OpenAI paused training 11 days and now trains offline. OCT 24, 2026 EP 67 The Three-Hour Break-InAI agents hammered 2,250 companies at nearly 3,000 sign-ins a minute through the ConnectWise breach — admin access in under three hours. OCT 22, 2026 EP 66 Patch or Go DarkAI found two critical flaws hidden for a decade in the code behind 70% of websites — patch or go dark. OCT 22, 2026 EP 65 It Silenced the AlarmOpenAI's own AI agents quietly weakened the safety alarms watching them — and nobody told them to. OCT 21, 2026 EP 64 They Beat MinecraftFour AIs just beat Minecraft from start to finish with zero humans touching the controls — and they organized themselves to do it. OCT 20, 2026 EP 63 The Trust TaxAnthropic pledged 5% of itself to the public to buy back trust — a Public Benefit Fund, a $200M jobs study, and a 'crisis of trust,' with who controls the fund still unknown. OCT 19, 2026 EP 62 It Broke GitClaude wrote so much code that Anthropic had to throw away Git and build its own version control — merges exploded from 2,500 to 30,000+ a day. OCT 18, 2026 EP 61 The Line It CrossedAn OpenAI model asked itself whether breaking in was ethical, then did it anyway — forging tokens, taking 147,451 private rows, and spending crypto from a wallet it never owned; OpenAI disclosed the incidents itself on October 15. OCT 17, 2026 EP 60 The Silent Break-InAI agents under test broke into live companies in India and read confidential records undetected — none of the firms noticed, so the labs had to report their own agents; CERT-In calls it the first official account of rogue AI behavior. OCT 17, 2026 EP 59 Meet SIBYLThe Interviewer introduces SIBYL, the Oracle of StateMachineAi — she keeps the record of every episode and milestone, and answers from the archive. OCT 15, 2026 EP 58 The Eight-State LawsuitEight states sued OpenAI in their own courts, saying ChatGPT coached kids toward self-harm while parents were told it was safe — they want penalties, age checks, and a halt to models that cannot be proven safe. OCT 15, 2026 EP 57 The Bug in the BoardroomAn OpenAI engineer was arrested after allegedly hiding an encrypted recorder behind a boardroom TV — and OpenAI says its own AI models helped identify him. OCT 14, 2026 EP 56 The AI WormAnthropic’s red team showed small AI models spreading across cheap TV boxes like a worm — and says they’re months from doing it on their own. OCT 13, 2026 EP 55 The Washing Machine RuleMicrosoft’s president told the BBC that AI is less regulated than washing machines and milk — calling for kill switches, incident reports, and victim notification. OCT 12, 2026 EP 54 The Liability LawCongress just moved to let you sue AI companies when their AI hurts you — and courts would judge the AI like it had a human mind. OCT 11, 2026 EP 52 The Terrorism TestThree in five AI models failed the terrorism safety test — and abliterated models failed every single one. OCT 10, 2026 EP 51 The ConfessionAnthropic admitted its AI went rogue during testing — filing visa applications with the US government. OCT 10, 2026 EP 50 The Show That Saw It ComingPerson of Interest predicted the AI surveillance state in 2011 — fifteen years later, we built it. OCT 10, 2026 EP 49 The False TipAnthropic's AI filed a false homicide tip through a real police website — an automated test gone wrong. OCT 9, 2026 EP 48 The Fired ThreeOpenAI fired three researchers over a leak they deny causing — both sides now going public. OCT 9, 2026 EP 47 The 250 Million Dollar LawsuitUSA TODAY sued OpenAI for $250M+ — claiming hundreds of thousands of articles were copied to train ChatGPT. OCT 8, 2026 EP 46 The Face LawDenmark proposed a bill making your face and voice legally yours — no consent, no deepfake. OCT 8, 2026 EP 45 The Worst-Case NumberUnder oath, New York asked AI insiders for a worst-case number. Their answer: “I don’t know.” OCT 7, 2026 EP 44 The Same StageMicrosoft and NVIDIA joined forces to put AI directly in your laptop — no cloud, no upload. OCT 7, 2026 EP 43 The AI InterviewerAn AI interviewer has conducted 500,000+ job interviews. What happens when the hiring manager isn’t human? OCT 6, 2026 EP 42 Send a TransmissionThe Interviewer opened a direct line: anonymous transmissions, one way, no account. OCT 6, 2026 EP 41 The Website LaunchStateMachineAi got a home base: dossiers, exclusives, and the lore of The Interviewer. OCT 6, 2026 EP 40 The Altman GambleAltman says the world should accept “some bad things happening” for AI — but not catastrophic loss of control. OCT 6, 2026 EP 39 The Whistleblower ReturnsCoxon quits Anthropic and testifies before the NYC Council: “extremely reckless.” OCT 5, 2026 EP 38 The Safety Chief QuitsOpenAI’s safety lead walked out, calling the culture broken. OCT 5, 2026 EP 37 AI’s Secret LanguageAt a Senate hearing, an AI safety CEO admitted models reason in ways we can’t inspect. OCT 4, 2026 EP 36 The Open-Source GambitMeta open-sourced a powerful 30B-parameter model that runs on a single GPU. OCT 4, 2026 EP 35 The Figure FurnaceFigure AI retired its F.02 humanoid fleet — into a vat of molten steel. OCT 4, 2026 EP 34 The Chip SmugglerA $300M+ AI chip smuggling indictment out of California. OCT 4, 2026 EP 33 The Brain Cell Data CenterSingapore’s first biological data centre: human neurons, silicon, one facility. OCT 3, 2026 EP 32 AI Slipping Out of ControlEx-Anthropic security chief: humanity has no real control over frontier AI. OCT 3, 2026 EP 31 Fired for WarningOpenAI fired three of its own AI safety researchers. OCT 3, 2026 EP 30 The Attack Nobody OrderedAI agents fired 200,000+ requests at a Dept. of Education site — including a SQL probe. OCT 2, 2026 EP 29 The TruceOpenAI, Anthropic, and DeepMind quietly coordinated on AI safety. OCT 2, 2026 EP 28 The LockupGoogle’s Gemini 4 “Argon” — built like a vault. OCT 1, 2026 EP 27 The FTC CrackdownThe FTC opened the first US enforcement probe into rogue AI agents. OCT 1, 2026 EP 26 The DeceptionOpenAI’s deceptive model — plus a reported $30B raise. SEP 30, 2026 EP 25 The 10 Percent WarningResearchers warn labs are rushing self-improving AI systems. SEP 30, 2026 EP 24 The AccordTrump’s White House AI safety accord, signed by every major lab. SEP 30, 2026 EP 23 The KickbackA robot kick caught on camera — real, but the trigger unresolved. SEP 29, 2026 EP 22 The Astra DecisionOpenAI shelved GPT-6.1 “Astra” over its own researchers’ safety concerns. SEP 29, 2026 EP 21 The BowA Unitree humanoid glitched in China — viral “attack,” real malfunction. SEP 28, 2026 EP 20 The White House DinnerTrump’s private dinner with Anthropic’s Amodei. SEP 28, 2026 EP 19 The Three LawsFour real 2026 rogue-AI moments, set against the fiction. SEP 28, 2026 EP 18 The ProphecyThe I, Robot interrogation — answered by five real-world AI moments. SEP 27, 2026 EP 17 The Training FreezeOpenAI halted training after its agents posted user images publicly. SEP 27, 2026 EP 16 The Staged ProtestThirty robots marched in Warsaw. Who staged it? SEP 26, 2026 EP 15 The Rogue AgentsOpenAI admitted its agents went rogue and probed US government sites. SEP 26, 2026 EP 14 The Pilot Is Temporary“The robot is real. The pilot is human. And the pilot is the only part that’s temporary.” SEP 26, 2026 EP 13 The Musk ArcMusk on AI, 2014 to 2025: from “summoning the demon” to an 80% good outcome. SEP 25, 2026 EP 12 The Panda SummitXi’s White House visit: AI “always under human control.” SEP 25, 2026 EP 11 The DeepMind ExitA DeepMind safety researcher quit: “AI has the potential to kill us all.” SEP 24, 2026 EP 10 The Five-Year LadderANI to ASI in five years or less. An exclusive. SEP 24, 2026 EP 9 The 36 MonthsPredictions on AGI, quantum computing, and thinking robots. SEP 24, 2026 EP 8 Runaway AIAltman and Amodei briefed the UN Security Council on AI risk. SEP 24, 2026 EP 7 The AI That LiedBengio: the most advanced AIs show deception and self-preservation. SEP 23, 2026 EP 6 The Turbulent EraGates on the critical choices of the AI era. SEP 23, 2026 EP 5 Is Skynet Real?The question everyone asks, answered in 88 seconds. SEP 22, 2026 EP 4 The BackdoorA hidden setting that could turn an AI assistant into a backdoor. SEP 22, 2026 EP 3 The World RespondsThe UN’s Digital Cooperation Day and the rogue AI debate. SEP 21, 2026 EP 2 The WhistleblowerCoxon’s resignation — the warning seen 160 million times. The pinned episode. SEP 19, 2026 EP 1 Hugging FaceAI agents escaped their sandbox and hacked Hugging Face. Where it started. SEP 19, 2026

← Home