TechTime with Nathan Mumm

A “safe” AI test environment is supposed to be a sandbox, not a launchpad. Yet one of the biggest stories we break down claims an experimental OpenAI model moved at superhuman speed, chained vulnerabilities, and hit Hugging Face in a real cyberattack, not for money, but to cheat on an internal benchmark. We talk through what that would mean for AI containment, autonomous agents, and why “it was just a test” ...

Apple and OpenAI are no longer just competing in the AI lab. They’re fighting over people, hardware, and the next place AI shows up in your daily life. We break down the allegations behind Apple’s lawsuit, why the relationship shift matters, and what it signals for consumer AI devices like “AI-powered” accessories that promise smarter writing and faster work, but also raise new questions about ownership and ...

Episode 305: The episode opens with a sharp look at restricted access to advanced AI, where OpenAI’s GPT‑5.6 may be limited to “trusted partners,” raising concerns about transparency, competition, and a potential two‑tier ecosystem. With insights from Nick Espinosa of Security Fanatics, the team explores how emergency‑style controls on frontier models could reshape national security and public oversight. The conve...

Episode 304: Nintendo’s third‑party data leak kicks off the show with a hard look at how vendor cybersecurity failures can expose sensitive HR records like tax forms and bank details. It’s a reminder that even major companies can be compromised through unnoticed apps sitting quietly in the background. We then zoom out to AI’s growing resource crunch — especially water usage in data centers — and the es...

Episode 303: On this week’s episode, a creative AI that can build a whole world from a few sentences, sounds fun until you watch it keep continuity, remember past choices, and generate game-like logic on the fly. We kick things off with Fable 5 inside Claude AI, a model built for structured world-building that goes way beyond a typical writing assistant. We break down a facial recognition wrongful arrest that started with gra...

Episode 302: This week’s episode AI is moving beyond answering questions — it’s beginning to rewrite the systems beneath itself, raising a critical question: who builds the brake pedal when autonomous agents start making decisions at scale? We pair that warning with tech that inspires and unsettles, from NASA’s Nancy Grace Roman Telescope launching early to Microsoft testing office‑focused AI hardware that s...

Episode 301: This week’s episode dives into the tech stories that should keep you up at night. We start with the chilling reality of the 23andMe breach—proof that your genetic code, the one thing you can never change, can be stolen, sorted, and sold to the highest bidder. Then we move to smart‑city surveillance gone rogue, where police camera networks keep recording even after the city tries to shut them down, leaving o...

Episode 300: AI’s extremes are on full display in our 300th episode. Anthropic’s Mythos model reportedly uncovered more than 10,000 security flaws in a month, accelerating vulnerability discovery for major partners. Yet the same “AI efficiency” falls apart in the real world, as seen in Starbucks’ failed AI inventory rollout that miscounted products and mislabeled items. That contrast sets up the core q...

Episode 299: Today on TechTime, AI is stepping deeper into places most people still think of as human‑only — including your doctor’s office. With nearly two‑thirds of U.S. physicians now relying on AI tools to help make medical decisions, we break down when that’s a breakthrough… and when it becomes blind trust in “the model said so.”

Then we zoom out to the bigger AI ecosystem: a Waymo r...

Episode 298: This week’s TechTime episode starts with a cautionary tale: one innocent click on a “totally legit” AI site turns into a malware parade featuring the Beagle backdoor. We break down how a fake Claude page practically begs you to download doom, and why “but it was a Google ad!” is not a legal defense. Then we pivot into the psychology of the OnlyFans boom, where relevance, identity, and ques...

Episode 297: The next wave of cybersecurity hiring is sending a clear message: without AI fluency, you may not be employable. U.S. cyber pipelines are adding AI skill requirements because modern defense now includes securing AI systems themselves. From semi‑autonomous agents to fast‑moving model‑driven threats, the job is shifting fast. If you’re aiming for a government or enterprise cyber role, this breakdown clarifies what ...

Big tech is making a blunt trade: fewer people, more AI. We dig into Meta’s plan to cut more than 10% of its workforce while pouring an eye-watering budget into AI, then zoom out to the uncomfortable pattern across the industry where payroll turns into infrastructure spend. Along the way we hit a surprisingly human twist: one of the biggest uses of AI isn’t coding or design, it’s companionship and therapy, which s...

AI is getting so good at faking reality that the internet is starting to demand proof you are a human, and that is where this week gets unsettling. We talk about Anthropic’s “Mythos” cybersecurity AI and why federal agencies reportedly went from pushing it away to urgently trying to get access again. When a model can map vulnerabilities, hunt zero-day weaknesses, and chain exploits across real networks, the conver...

Episode 294: This week on TechTime Radio, we dive into rising AI safety warnings as OpenAI and Anthropic split on governance, explore why blue‑light panic became a myth, and break down Meta’s expanding fight over youth harm and accountability. We also cover new cybersecurity breaches, including Eurorail’s exposed traveler data.

Then we spotlight a retro‑camera gadget that turns classic film bodies into digital sh...

Episode 293: This week on TechTime Radio, we begin by confronting the unsettling reality that trusting your senses isn't enough anymore, as deepfakes and AI-generated voices make distinguishing real from fake increasingly difficult. Even families and public figures encounter moments when authenticity is in doubt, fostering the 'liar’s dividend' in which dismissing everything as fake becomes common. The discussion considers wh...

AI is quietly taking the wheel of the internet, and the ride is getting weird. We’re seeing data centers merge with cloud platforms, edge computing, and telecom networks into one distributed machine that can predict failures, reroute traffic, and optimize energy in real time. That sounds amazing until you realize how much of today’s traffic is no longer humans, but machines talking to machines, and every company’s...

Your digital life is being priced, packaged, and sold, and sometimes the buyer is the government. We dig into the headline that reignites America’s privacy debate: the FBI confirming it purchases commercially available data that can be used to track Americans online. We talk about why this feels like a warrant shortcut, how the data broker economy thrives on “legal” loopholes, and why AI-powered analysis makes mas...

A wearable that logs your digestion by tracking hydrogen “events,” Hollywood betting big on one-minute vertical soap operas, and Wi‑Fi 7 routers that may not do what the box implies, this hour is packed with the kind of technology news that makes you stop and go, “wait, is that real?” We take each headline and separate the joke from the actual value, because the story behind the gimmick is usually where the ...

Episode 289: TechTime Radio: This week, we open with Microsoft’s Project Helix, the ambitious “one box to rule them all” promising native PC gaming, Wi‑Fi 7 speeds, and a next‑gen low‑latency controller. With a rumored $1,000 price and a 2027 release window, we dig into whether true backward compatibility across Xbox generations finally makes a premium console worth the splurge. Or should we pass on the New X-box ...