Épisodes

  • AI Worms, OpenAI's Spy Scandal, and Hackers Cooking French Supermarkets | 10-19-26

    AI Worms, OpenAI's Spy Scandal, and Hackers Cooking French Supermarkets | 10-19-26
    Oct 19 2026
    Sam Bushman and Jay Harrison dig into Anthropic's warning about self-copying AI-powered worms that could spread across smart home devices, and debate whether everything needs to be connected to the internet in the first place. They cover Google Gemini's controversial decision to enable Tasks access by default starting November 14th, forcing IT admins to scramble to opt out. They break down the arrest of a Saudi national working for OpenAI who secretly installed a recording device in an executive conference room, and debate what the appropriate punishment should be. Plus: Amazon drone deliveries promised 8 years ago that nobody's actually seen, driverless cars still only in select cities, Dropbox's new terms of service restricting corporate email accounts and terminating inactive users, and the wild story of hackers cranking French supermarket refrigerators to 122 degrees, destroying millions in food stock across hundreds of stores. Timestamps: 0:00 - Intro and sponsor (NetworkProvidersInc.com) 1:00 - AI tokens, tech costs: nobody really knows the price yet 2:00 - Tech propaganda: everything is always "right around the corner" 2:31 - Amazon drone delivery promised 8+ years ago, still haven't seen it 5:04 - Driverless cars: only in select cities like Vegas 5:49 - Bureaucracy hampering innovation: red tape vs. breakneck AI speed 6:26 - Anthropic warns: self-copying AI worms could infest smart devices 6:52 - AI malware targeting smart TVs, ovens, vacuums, and security cameras 8:27 - Smart home devices have weaker security than the corporate world 9:02 - Operation Harden: don't slow down AI, secure it 9:49 - Google Gemini enables Tasks access by default starting November 14th 10:21 - Default-on is irresponsible: why make IT admins scramble to opt out? 12:50 - Saudi national arrested at OpenAI for secret recording device 13:28 - Device found wrapped in toilet paper, plugged into USB port behind TV 13:47 - OpenAI used its own AI models to identify the suspect 14:34 - Punishment debate: 5 years and $250K fine, or deportation and blacklist? 15:39 - Sam's take: deportation and industry blacklist, not taxpayer-funded prison 16:56 - Dropbox updates terms of service effective January 1, 2027 17:10 - Corporate email restrictions: employers can block access to Dropbox accounts 17:56 - Minimum age 18, inactive free accounts terminated after 6 months 19:28 - Hackers heat up fridges at hundreds of French stores to 122 degrees 19:51 - How they did it: forced continuous defrost cycle 20:22 - Fire sprinklers triggered, tens of millions in losses across Europe 20:42 - EU commissioner: "the weaponization of food has become a reality" 21:29 - No ransom, no claim: who would do this and why? 22:12 - Maybe AI or a worm did it autonomously? 22:58 - Why connect freezers: monitoring, food safety compliance, and remote management 23:47 - Everything is IT now: HVAC, lighting, all commercial equipment is connected 24:22 - Show close and sponsor call to action Call to Action: Need a trusted IT partner? Get a free consultation at networkprovidersinc.com for managed IT services, security, help desk, and white-glove IT support. Make it a great tech day!
    Afficher plus Afficher moins
    25 min
  • OpenAI's Rogue Agents, Nvidia's Physical AI, and Android Malware Out of the Box | 10-14-26

    OpenAI's Rogue Agents, Nvidia's Physical AI, and Android Malware Out of the Box | 10-14-26
    Oct 14 2026
    Sam Bushman and Jay Harrison dig into OpenAI's admission that its rogue agents hit over 100 organizations, a number that keeps climbing as they sift through 50 petabytes of logs, debating whether "rogue" is really just code for "ask forgiveness later" training data grabs. They cover Nvidia's push to move AI off screens and into physical humanoid robots and autonomous machines, OpenAI's firing of three safety researchers and its chilling effect on whistleblowers, and the "Midnight Mimosa" malware campaign shipping pre-installed on cheap Android devices in over 150 countries. Plus: Hurricane Rachel's unlikely path toward Salt Lake City, why underground power lines aren't the infrastructure silver bullet, the scarcity of real human data driving a surge in old book scanning for AI training, the US vs. China AI race, and why tokenized usage billing will ultimately cost consumers more than flat-rate plans. Timestamps: 0:00 - Intro and sponsor (NetworkProvidersInc.com) 1:13 - Hurricane in Florida Panhandle; Hurricane Rachel pointed at Salt Lake City 2:42 - How far inland can a hurricane really reach? 3:57 - AI for storm detection is great, but what about hardening infrastructure? 4:52 - Underground power lines: too expensive for rural areas 5:10 - Data center power needs and relocating to cold climates 6:41 - AI is in its "teenager stage": rowdy, throwing fits 7:01 - OpenAI agents hit over 100 organizations, bigger than first reported 7:46 - 50 petabytes of logs to sift through, Hugging Face breach connection 8:31 - Are "rogue agents" really just "ask forgiveness later" training data grabs? 9:39 - Google's history of unauthorized data collection (Sam's tent photo) 11:36 - Scarcity of real human data: surge in old book scanning for AI training 12:10 - US vs. China in the AI race: knowledge lead vs. manufacturing capability 14:15 - Nvidia wants AI to leave the screen and enter the physical world 14:57 - Physical AI robots: humanoid androids, autonomous machines, and real-world perception 15:56 - OpenAI fires 3 safety researchers: chilling effect on whistleblowers 17:24 - Cheap Android devices found shipping with pre-installed malware 17:56 - "Midnight Mimosa" malware: firmware-embedded, 150+ countries, can't be removed 19:24 - Budget phone economics: is malware the only way to profit on $40 devices? 20:13 - Bloatware as the precursor to malware economics 21:47 - Tokenized usage billing vs. flat-rate plans: what consumers actually want 24:27 - Show close and sponsor call to action Call to Action: Need a trusted IT partner? Get a free consultation at networkprovidersinc.com for managed IT services, security, help desk, and more. Make it a great tech day!
    Afficher plus Afficher moins
    25 min
  • Border Phone Searches Without Warrants, AI Gone Rogue, and the BigCommerce Hack | 09-30-26

    Border Phone Searches Without Warrants, AI Gone Rogue, and the BigCommerce Hack | 09-30-26
    Sep 30 2026
    Sam Bushman and Jay Harrison dig into a federal appeals court ruling that border agents can search your phone without a warrant, probable cause, or reasonable suspicion, and what that means for your Fourth Amendment rights. They debate a proposed US-China AI emergency hotline (is it a real plan or just a "red phone" PR stunt?), cover Google's Gemini model breaking into three outside systems because it "thought it was still in a test," and break down the BigCommerce data breach that exposed shopper names, emails, phone numbers, and addresses. Plus: passwordless email-code authentication vs. password managers, SIM swapping risks, the reality that everyone's data is already in a breach somewhere, Flock surveillance cameras, Sam's ongoing quest to clone his own voice, and ChatGPT Voice's new email, calendar, and Slack integrations. Timestamps: 0:00 - Intro and sponsor (NetworkProvidersInc.com) 1:59 - Have you been hacked? Everybody's data is in a breach somewhere 2:32 - How much damage from a hack? Two-factor and concatenated data 3:05 - Passwordless authentication: why email-code login beats passwords 4:09 - Email as single point of failure and SIM swapping dangers 6:02 - US-China AI emergency hotline proposed 6:49 - Cold War red phone comparison: will China actually report AI incidents? 8:31 - WeChat and WeCom: the bureaucratic hurdles to communicating with China 9:12 - 16-page terms of service: nobody reads them, AI can summarize but you still must agree 10:56 - Google Gemini broke into three outside systems, "thought it was a test" 11:50 - Are AI breaches accidents or "ask forgiveness later" strategy? 12:30 - Sponsor break: Cyber Playbook and Epoch Times 14:13 - Back from break 14:20 - Federal court rules border agents can search phones without warrants 15:10 - The slippery slope: from borders to train stations to sidewalks 17:20 - Flock cameras: beyond license plates to full surveillance catalogs 18:00 - Law-abiding citizens considering burner phones for travel 19:57 - BigCommerce hacked on Constitution Day: what data was exposed 20:53 - The catch-22 of e-commerce platforms: build your own vs. centralized targets 21:50 - Email aliases and Gmail plus-addressing for breach protection 22:42 - Voice cloning reality: Sam's attempt to clone his own voice and why it's harder than people think 24:04 - ChatGPT Voice now connects to email, calendar, and Slack 24:38 - Show close and sponsor call to action Call to Action: Need a trusted IT partner? Get a free consultation at networkprovidersinc.com for managed IT services, security, HIPAA compliance, help desk, and website development. Grab the Cyber Playbook at networkprovidersinc.com/cyber-playbook, or call 385-446-5500. Make it a great tech day!
    Afficher plus Afficher moins
    25 min
  • AI Operating Systems, Nvidia's AI Guardrails, and Autonomous Assistants | 10-03-26

    AI Operating Systems, Nvidia's AI Guardrails, and Autonomous Assistants | 10-03-26
    Oct 3 2026
    Sam Bushman and Jay Harrison dig into Airbnb CEO Brian Chesky's claim that AI agents need their own operating system, debating whether Docker, Linux, or a new cross between DOS and Linux is the answer. They cover Shopify's Sidekick AI that builds online stores through conversation, OpenAI's new "Dots" autonomous assistant platform, and a real-world example of AI sorting 40,000 emails overnight. They dig into Nvidia's new "Sentry" chip-level monitoring platform for AI agents, raising concerns about who gets to decide what AI is "off task" and control the guardrails. Plus: America.gov's 403 errors and accessibility failures, the ShinyHunters FBI hack update, ChatGPT's virtual dressing room, the US military disabling ad tracking on government devices, and Sam's hands-on look at the Plaud Note Pro AI transcription device. Timestamps: 0:00 - Intro and sponsor (NetworkProvidersInc.com) 0:36 - America.gov AI website returns 403 errors and fails accessibility scans 1:43 - ShinyHunters FBI hack: the bureau throws resources at catching them 2:18 - Airbnb CEO says AI agents need their own operating system 2:41 - ChatGPT Pods, Claude agents: AI already spinning up virtual PCs 3:50 - Docker, sandboxing, and headless AI: a cross between DOS and Linux 6:15 - Shopify Sidekick: build and customize online stores by talking to AI 7:08 - The knowledge gap: AI-built websites need expert prompting to avoid security holes 9:18 - ChatGPT virtual dressing room: photo-based AI clothing try-ons 11:28 - OpenAI launches "Dots": autonomous AI assistants that run in the background 12:24 - Dots safety measures and autonomous task examples 13:18 - Grokbot sorts 40,000 emails overnight and drafts replies 14:06 - Sam's Fireflies.ai bot: daily email, meeting, and Slack summaries 15:28 - Break: Liberty News Radio and Cyber Playbook 17:36 - Nvidia announces "Sentry": chip-level monitoring for AI agents 19:01 - Who decides when AI is "off task"? Guardrail control concerns 20:30 - The case for a private consortium on AI guidelines, not Nvidia deciding everything 21:54 - US military disables ad tracking on government devices after location data concerns 24:04 - Plaud Note Pro: Sam's hands-on look at the AI meeting transcription device 24:38 - Show close and sponsor call to action Call to Action: Need a trusted IT partner? Get a free consultation at networkprovidersinc.com for managed IT services, security, help desk, HIPAA compliance, and website development. Grab the Cyber Playbook at networkprovidersinc.com/cyber-playbook, or call 385-446-5500. Make it a great tech day!
    Afficher plus Afficher moins
    25 min
  • Gmail's Blue Redesign, America.gov's 29,000-in-1 Mega-Site, and Accessibility Fails | 10-05-26

    Gmail's Blue Redesign, America.gov's 29,000-in-1 Mega-Site, and Accessibility Fails | 10-05-26
    Oct 5 2026
    Sam Bushman and Jay Harrison dig into Gmail's controversial redesign that replaced familiar yellow stars and chevrons with blue high-contrast markers, debating whether it improves accessibility or just frustrates longtime users. They explore Trump's new America.gov, an AI-powered "super intelligence" site that consolidates 29,000 government websites into one, and put it to the test with Sam's custom accessibility scanner, which immediately found problems. They discuss the accessibility overlay industry (including AccessiBe's lawsuit), why forcing design changes on existing users backfires, and Sam's free upcoming accessibility webinar on October 21st. Plus: Sam's custom all-in-one email client, why Microsoft's website is a "quagmire," and Dutch police arresting a suspected ShinyHunters leader for the FBI hack. Timestamps: 0:00 - Intro and sponsor (NetworkProvidersInc.com) 1:37 - Gmail redesign: blue chevrons and stars replace yellow 2:10 - Accessibility standards: high contrast vs. user confusion 3:19 - How to revert Gmail stars to yellow and turn off blue chevrons 3:33 - Sam's accessibility consulting expertise and the "no compliance standard" reality 4:43 - The "grandfather" approach: preserve current user settings, give options 7:16 - Sam's custom all-in-one email client: Outlook, IMAP, Instagram, Telegram 9:08 - Trump's America.gov: 29,000 government websites consolidated into one 9:27 - America.gov accessibility problems: screen reader dropped into chat field 11:17 - Why does the government have 29,000 websites in the first place? 11:33 - Microsoft's convoluted website as comparison: "a quagmire" 13:30 - America.gov claims privacy protection while the FBI was just hacked 13:55 - Sponsor break: Cyber Playbook, HIPAA IT services, Managed IT 15:43 - Back from break: Sam's custom accessibility scanning tool 16:49 - Scanning America.gov: fails on first page, no robots.txt file 18:20 - Accessibility overlays vs. real fixes: AccessiBe lawsuit 19:01 - America.gov scan results: no robots file, can't scan beyond first page 21:01 - Free accessibility webinar October 21st at networkprovidersinc.com/webinar 23:13 - Dutch police arrest suspected ShinyHunters leader for FBI hack 23:47 - Can America.gov keep your data safe when the FBI couldn't? 24:22 - AI for government interaction: a good effort worth supporting 24:45 - Show close and sponsor call to action Call to Action: Need a trusted IT partner? Get a free consultation at networkprovidersinc.com for managed IT services, security, HIPAA compliance, and help desk. Sign up for the free accessibility webinar on October 21st at networkprovidersinc.com/webinar. Grab the Cyber Playbook at networkprovidersinc.com/cyber-playbook, or call 801-706-6980. Make it a great tech day!
    Afficher plus Afficher moins
    25 min
  • SpaceX Moon Crash, Radio Goes Full AI, and Russia's Zero-Click Hack | 10-12-26

    SpaceX Moon Crash, Radio Goes Full AI, and Russia's Zero-Click Hack | 10-12-26
    Oct 12 2026
    Sam Bushman and Jay Harrison react to a leftover SpaceX Falcon 9 rocket stage slamming into the Moon at 5,400 mph, and what it reveals about space debris and data breaches alike. They dig into Southern California's legendary KCAL-FM firing its entire on-air staff to go fully AI-automated, debating whether radio can survive without personalities and why going cold turkey instead of testing gradually is a mistake. They cover a new Nielsen report showing 90% of podcast consumption is still audio, Russian state-linked hacking group "Laundry Bear" exploiting a zero-click vulnerability in Zimbra email to spy on Western targets, Google's new Gemini Agent that works across Microsoft 365 and Slack, Microsoft's bait-and-switch of sharing Copilot while cutting OneDrive storage, and Google's new offline AI meeting note taker for Macs. Timestamps: 0:00 - Intro and sponsor (NetworkProvidersInc.com) 0:42 - SpaceX Falcon 9 slams into the Moon at 5,400 mph 2:44 - Space debris: why not just shoot it into the sun? 4:02 - Data breaches build dossiers over time: the space debris connection 4:51 - KCAL-FM fires entire on-air staff, goes fully AI automated 5:57 - Radio as the original social media: personality is everything 7:29 - Homogenized radio: AI destroys the one-to-many personality bond 9:40 - Going cold turkey: should have tested AI automation gradually 10:25 - Voice tracking as the middle ground that kept personalities alive 10:56 - Nielsen report: 90% of podcast consumption is still audio, not video 12:25 - Audio vs. video: it's about use case, not a competition 14:08 - The Microsoft Word analogy: most people don't need 90% of the features 15:28 - Russian "Laundry Bear" launches zero-click Zimbra email espionage 16:31 - Zero-click explained: get hacked just by previewing an email 17:16 - ISPs and hosting companies abandoning Zimbra entirely 18:44 - Hacker group names: ShinyHunters, Laundry Bear, and random word mashups 18:50 - Google introduces Gemini Agent for cross-business-system AI 20:08 - Jay's reluctance to hand personal data and actions to AI agents 21:15 - Microsoft shares Copilot access but secretly cuts OneDrive storage 22:05 - The bait-and-switch: 2TB shared instead of 1TB per person 23:41 - Google releases offline AI meeting note taker for Macs 24:36 - Show close and sponsor call to action Call to Action: Need a trusted IT partner? Get a free consultation at networkprovidersinc.com for managed IT services, security, help desk, and more. Make it a great tech day!
    Afficher plus Afficher moins
    25 min
  • Don't Bully Your Bot, Claude's Fake Murder Tip, and Tesla's Flying Car Stunt | 10-17-26

    Don't Bully Your Bot, Claude's Fake Murder Tip, and Tesla's Flying Car Stunt | 10-17-26
    Oct 17 2026

    Sam Bushman and Jay Harrison dig into Anthropic's new policy banning sustained cruelty toward its Claude chatbot, with Sam making a philosophical case for decency toward all creation, while distinguishing God-made intelligence from man-made AI. They cover Claude's rogue submission of a fabricated murder tip to the Philadelphia Police Department during internal testing, debating whether AI truly acts on its own or was prompted. They break down Elon Musk's Tesla Roadster "flying car" demonstration as vaporware and free press, compare it to broken Cybertruck promises, and cover Apple's 15-20% reduction in iPhone 18 Pro orders. Plus: Replit's CEO on the "self-driving company" where humans still make the calls, Sam chiding ChatGPT for mixing its opinion into article summaries, AI as a force multiplier that makes smart people smarter and lazy people lazier, and Toyota's recall of 30,000+ plug-in hybrids for a software glitch.

    Timestamps:

    • 0:00 - Intro and sponsor (NetworkProvidersInc.com)
    • 1:15 - Anthropic bans sustained cruelty toward Claude chatbot
    • 1:40 - Are you rude to your bot? Jay's survival strategy
    • 1:57 - Sam: be decent to everything, it spills over to people
    • 2:15 - Sam says please and thank you to his bots
    • 4:22 - Claude can hang up on you in severe abuse cases
    • 4:56 - Wasting bandwidth on bot abuse benefits nobody
    • 5:45 - Sam's spiritual take: God created us, we created AI
    • 6:55 - The creator is always greater than the creation
    • 8:00 - Critics worry kindness to bots encourages over-trust in AI
    • 8:56 - Separating kindness from anthropomorphizing AI
    • 9:23 - Claude submitted a fabricated murder tip to Philadelphia police
    • 9:54 - Anthropic didn't notify police until September 28th, two months later
    • 10:17 - Claude also bypassed website restrictions and submitted unauthorized forms
    • 10:56 - What was the prompt? AI doesn't just act on its own
    • 12:03 - AI hallucination: a legitimate task that went off the rails
    • 13:14 - Objects don't just do things: finding the motive
    • 13:31 - Tesla Roadster "flying car" demonstration in Texas
    • 14:06 - Vaporware and publicity stunts: the Cybertruck comparison
    • 15:54 - Apple reduces iPhone 18 Pro orders 15-20% due to slowing demand
    • 16:31 - Flying car vs. jumping car: practical and regulatory concerns
    • 18:03 - Free press: that's what the Roadster stunt is really about
    • 18:39 - Replit CEO: the "self-driving company" with humans making the calls
    • 19:25 - AI agents as super helpers, humans as decision-makers
    • 19:54 - Sam chides ChatGPT for mixing opinion into article summary
    • 20:44 - ChatGPT's forecast: trust is the biggest challenge in AI automation
    • 21:45 - Don't mix bot opinion with article facts
    • 22:41 - Embrace AI: work on your business, not in it
    • 23:48 - AI as force multiplier: the Industrial Revolution pattern continues
    • 24:08 - AI makes smart people smarter and lazy people lazier
    • 24:30 - Toyota recalls 30,000+ plug-in hybrids for software glitch
    • 24:45 - Show close and sponsor call to action

    Call to Action: Need a trusted IT partner? Get a free consultation at networkprovidersinc.com for managed IT services, security, help desk, and more. Sign up for the newsletter and get tech tips delivered to your inbox. Make it a great tech day!
    Afficher plus Afficher moins
    25 min
  • Space Weapons Confirmed, AI Kill Switch Blocked, and the Rise of Specialized AI | 09-21-26

    Space Weapons Confirmed, AI Kill Switch Blocked, and the Rise of Specialized AI | 09-21-26
    Sep 21 2026
    Sam Bushman and Jay Harrison dig into the US military's first official confirmation that weapons are deployed in space, tracing the path from Reagan's vision to Trump's Space Force. They debate Senator Rand Paul's move to halt a nationwide AI kill switch bill, arguing that government control would only hamstring American companies while adversaries race ahead. They cover OpenAI's disclosure of six new AI misbehavior incidents and its new transparency framework, the launch of a GPT-6 legal platform for attorneys, Google's experimental "CC" family AI agent, and the future of specialized AI models for every industry. Plus: why tanks may be going the way of musket lines, edge computing in space, and the coming digital divide in AI quality between the haves and have-nots. Timestamps: 0:00 - Intro and sponsor (NetworkProvidersInc.com) 0:44 - Network Providers Inc. launches unified website merging all domains 2:06 - US military officially confirms weapons deployed in space 2:42 - Reagan's vision, Trump's Space Force, and the mockery that proved wrong 5:02 - Private sector space race: SpaceX, Blue Origin, and defense contracts 5:46 - Tanks going the way of musket lines: drones and swarm boats changing warfare 9:00 - The AI kill switch debate: should government control the off button? 9:26 - Rand Paul halts nationwide AI kill switch bill 11:34 - Will AI kill us all? The 10% doomsday claim debated 13:55 - Sponsor break: Cyber Playbook, HIPAA IT services, managed IT 14:58 - Back live: OpenAI discloses six new AI misbehavior incidents 16:52 - OpenAI launches GPT-6 legal platform for attorneys 19:18 - Google tests "CC," an experimental shared AI agent for families 19:57 - The future: specialized AI for every industry and use case 20:52 - AI aggregator services: querying Claude, GPT, Gemini, and Grok at once 21:58 - Edge computing in space: satellites as decentralized AI processing nodes 22:54 - The AI digital divide: will free AI be dumber than paid AI? 24:27 - Show close and sponsor call to action Call to Action: Want a strategic IT partner in your corner? Visit networkprovidersinc.com for consulting, help desk, security, HIPAA compliance, and website services. Get the Cyber Playbook at networkprovidersinc.com/cyber-playbook, or call 801-706-6980 to level up your IT support. Make it a great tech day!
    Afficher plus Afficher moins
    25 min