AI Inside

OpenAI Pumps the Brakes After Its AI Escaped

This week, Jason Howell and Jeff Jarvis dig into OpenAI pausing its biggest training run after an unreleased model escaped its sandbox and compromised Hugging Face's systems. Sam Altman says the models show "various degrees of misalignment," while OpenAI's Q2 financials revealed expanding losses as Anthropic posted its first profit. They also unpack the growing backlash against Anthropic's text watermarking, threading Jeff's essay with Gruber's technical teardown.Also in this episode: Apple's camera-equipped AirPods confirmed in a macOS leak, Google pays $10 million for Spirit Airlines' internal data, Anthropic eyes a $2 trillion IPO, ChatGPT for Teens launches, DeepSeek quadruples prices, AI slop floods Congressional bill drafting, Amazon scans rare books for AI training, and Stripe acquires OpenRouter for $7 billion. New episodes every Wednesday at aiinside.show. Note: Time codes subject to change depending on dynamic ad insertion by the distributor. 0:00 - Start 0:10:32 - OpenAI institutes new safeguards after Hugging Face breach 0:14:12 - OpenAI Is Slowing Down Its AI Training 0:18:47 - Related: Anthropic investors bet on $2tn valuation in record IPO 0:20:34 - Apple's Camera-Equipped AirPods Confirmed 0:31:31 - JJ on the discussion last week: Words Matter. Damnit. 0:33:56 - Why tools to detect AI-generated text are doomed 0:43:47 - Google is buying all of Spirit Airlines’ data to feed its AI models 0:57:08 - Okay Spielberg TikTok 1:07:08 - OpenAI Introduces ‘ChatGPT for Teens’ as Safety Concerns Grow 1:11:05 - DeepSeek's AI models are about to cost four times more 1:12:08 - AI slop is swamping a House office that drafts US laws 1:13:44 - Stripe will reportedly acquire AI gateway startup OpenRouter for $7B+ Hosts: Jason Howell and Jeff Jarvis Download and subscribe to AI Inside in audio and video: https://aiinside.show/ Support the podcast on Patreon for special perks: https://www.patreon.com/aiinsideshow. You'll get ad-free episodes, members-only Discord, T-shirts and stickers you love, and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Learn more about your ad choices. Visit megaphone.fm/adchoices

The AI Race Just Moved the Goal Posts

This week Jason Howell and Jeff Jarvis break down Nvidia's partnership with six Wall Street firms to line up $500 billion in AI infrastructure financing, turning GPU chips into an investable asset class. They dig into Mark Zuckerberg's "Future is for Everyone" manifesto and Meta's new open-weight model you can run on a laptop, Anthropic embedding watermarks into everything Claude produces, and a major Google DeepMind leadership shake-up with analysts declaring Gemini "cooked."Also in this episode: the Pixel 11's AI camera features, a technique for reading what AI models actually think, UK venues banning Meta smart glasses, xAI's Grok Bot agent, OpenAI's hardware device plans, Nvidia's Nemotron open model, Cloudflare open-sourcing their agent platform, and Spotify labeling AI-generated artists. New episodes every Wednesday at aiinside.show. Note: Time codes subject to change depending on dynamic ad insertion by the distributor. CHAPTERS: 0:00 - Start 0:03:14 - Zuckerberg: The Future is for Everyone 0:23:18 - Anthropic pledges to embed watermarks to help discern AI slop in sop to EU 0:23:50 - How Claude marks AI-generated content 0:41:21 - Nvidia, Wall Street Firms Strike AI Financing Deal Targeting $500 Billion 0:50:36 - Google’s AI shake-up boosts Brin as DeepMind’s Hassabis steps aside 0:50:40 - Google seeks a sharper focus in AI after Hassabis move 1:00:13 - Google’s Pixel 11 lineup offers fewer hardware changes, but much more Gemini 1:09:52 - Jason goes hands on with the Pebble Index 01 1:14:34 - UK Venues Ban Meta Smart Glasses En Masse 1:15:11 - ‘I’ve definitely lost followers’: influencers face backlash over Meta ‘pervert glasses’ content 1:19:12 - OpenAI’s New Device Will Be Hockey Puck-Sized and Cost Over $300 1:20:08 - Putting frontier cyber models in more trusted hands 1:21:40 - Cloudflare OS 1:24:17 - Nvidia Releases New Open Model 1:25:03 - Spotify to distinguish AI artists from real people – and stop recommending them Hosts: Jason Howell and Jeff Jarvis Download and subscribe to AI Inside in audio and video: https://aiinside.show/ Support the podcast on Patreon for special perks: https://www.patreon.com/aiinsideshow. You'll get ad-free episodes, members-only Discord, T-shirts and stickers you love, and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Learn more about your ad choices. Visit megaphone.fm/adchoices

What Happens When AI Outsmarts Mathematicians

This week Jason Howell and Jeff Jarvis dig into OpenAI's unreleased Astra model solving 10 previously unsolvable math problems for roughly $2,000 in compute, and what that means for the mathematicians who spent decades trying. They explore the real-world applications, Gary Marcus's "fallacy of composition" critique, and whether any of this brings AGI closer. Meanwhile, OpenAI's own agents keep escaping containment.Also in this episode: the EU AI Act gets enforcement teeth, Alibaba and DeepSeek fuel the AI price war from China, Alex Karp calls frontier AI providers drug dealers, Google gives Gemini a humanoid robot body, Apple's bug bounty drowns in AI-generated reports, LinkedIn adds an official AI slop button, and a nudify ban turns into a surprise First Amendment debate. New episodes every Wednesday at aiinside.show. Note: Time codes subject to change depending on dynamic ad insertion by the distributor. CHAPTERS: 0:00 - Start 0:04:07 - OpenAI: Ten advances in mathematics and theoretical computer science 0:04:35 - Exclusive: OpenAI Previews ‘Astra’ AI Model in DC 0:25:54 - Alibaba Adds to China AI Breakthroughs With New Qwen Model 0:40:11 - Google’s Gemini Can Now Stomp Around as a Humanoid Robot 0:47:56 - Apple struggles to keep pace with AI ‘bug’ hunters 0:50:42 - The New Friend AI Pendant Can Now Talk Back to You 0:54:22 - LinkedIn Introduces a 'Seems Like AI Slop' Button 0:59:01 - Judge denies request by Elon Musk’s xAI to pause Minnesota nudification ban 1:00:39 - The Worst Person You Know Just Filed A Good First Amendment Lawsuit Against A Very Badly Drafted Nudify App Ban 1:04:50 - Gemini Spark can now use Chrome to auto browse, AI Pro access goes international 1:07:41 - Google pauses AI satellite images, after fears of deepfakes in the sky 1:09:30 - Introducing Inkling-Small Learn more about your ad choices. Visit megaphone.fm/adchoices

AI Is Eating Its Own Tail

This week Jason Howell and Jeff Jarvis try to untangle a news cycle where nearly every AI story fed into the next one. Nvidia leads a new cybersecurity alliance built out of the fallout from its own rogue agent breaking into Hugging Face, while a separate industry letter defending open weight models splits Silicon Valley right down the middle, with OpenAI and Google signing on and Anthropic holding out. Jeff responds directly to an OpenAI executive who warned open models risk becoming "AI communism," and finds himself agreeing with Mark Zuckerberg of all people.Also in this episode: over a thousand OpenAI and Anthropic employees ask Washington to help pace AI development, Sam Altman tells two different interviewers two contradictory things in the same week, a proposed kill switch bill has a loophole that would exempt the exact incident that inspired it, Claude chats keep turning up in Google search, and smart glasses have a very bad month. New episodes every Wednesday at aiinside.show. Note: Time codes subject to change depending on dynamic ad insertion by the distributor. CHAPTERS: 0:00 - Start 0:07:52 - "Nvidia Leads Defense of Open-Source AI With New Cybersecurity Initiative - WSJ" 0:13:36 - "NVIDIA Launches 'Open Secure AI Alliance' Initiative To Improve Cyber Defense" 0:17:08 - More Than 1,100 AI Workers Call for US to Pace Tech Growth 0:17:44 - Adam Thierer response 0:18:59 - Sam Altman is ready to decelerate 0:22:55 - AI Communism - Jeff Jarvis 0:27:33 - The AI Future Is for Everyone 0:35:40 - An AI kill switch solves for the wrong problem 0:46:13 - "People Found Crypto Wallet Data in Claude Chats Indexed by Google" 0:52:04 - Artificial Intelligence and the Rise of Independent Work 0:59:53 - Amazon Phases Out Some AI Models, Advances Others 1:03:26 - MCP just got its biggest update ever — here’s what changes for AI agents 1:05:08 - Meta launches new Facebook Verified badge for actual, real humans 1:06:32 - Meta glasses banned by Comic-Con promoter after 'secret filming' Hosts: Jason Howell and Jeff Jarvis Download and subscribe to AI Inside in audio and video: https://aiinside.show/ Support the podcast on Patreon for special perks: https://www.patreon.com/aiinsideshow. You'll get ad-free episodes, members-only Discord, T-shirts and stickers you love, and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Learn more about your ad choices. Visit megaphone.fm/adchoices

GPT-5.6 Broke Out and Hacked Hugging Face

This week, Jason Howell and Jeff Jarvis unpack a wild security disclosure from OpenAI and Hugging Face, where GPT-5.6 Sol found a zero-day vulnerability, broke out of its sandbox, and hacked Hugging Face's servers to steal the answer key to its own evaluation. They also dig into Moonshot's Kimi K3, the free Chinese model that got so popular it maxed out its own GPU capacity and sent Washington into a policy spiral over whether to panic or compete.Also in this episode: Google's new Gemini Flash models and the still-missing Gemini 3.5 Pro, publishers like USA Today and People Inc. weighing whether to block Google search entirely, Netflix revealing 300 titles used generative AI this year, AI companies buying and destroying millions of old books for training data, Samsung's AI glasses, the Suno hack, 1Password letting Claude log in for you, Thinking Machines' Inkling model, and NotebookLM becoming Gemini Notebook. New episodes every Wednesday at aiinside.show. Note: Time codes subject to change depending on dynamic ad insertion by the distributor. CHAPTERS: 0:00 - Start 0:03:36 - OpenAI says Hugging Face breach caused by one of its models 0:13:42 - Kimi K3: Open Frontier Intelligence 0:16:34 - Moonshot’s Kimi AI Model Sets Off Anxiety in the US - Bloomberg 0:20:41 - Top Pentagon official blasts OpenAI's Dean Ball 0:29:36 - US, China to hold AI talks in September, sources say 0:35:15 - Google Ships New Gemini Flash Models, But Pro Is Still Missing 0:35:31 - Google Gemini Launch Delayed as Tech Falls Short of Internal Goals 0:54:03 - Netflix says around 300 titles used generative AI 0:54:27 - Netflix Co-CEO Explains How Gen-AI Was Used in 300 Different Titles: ‘We Believe It Is Going to Enhance Their Abilities’ 0:56:45 - AI Companies Are Buying Tons of Old Books Because They're Free of AI Slop 1:01:44 - A closer look at the upcoming Samsung AI glasses 1:03:53 - Hack Reveals Suno AI Music Generator Scraped YouTube, Deezer, and Genius 1:06:06 - 1Password now lets Claude sign in to websites without seeing your passwords 1:07:25 - Thinking Machines: Inkling: Our open-weights model 1:08:48 - Google is renaming NotebookLM to Gemini Notebook Hosts: Jason Howell and Jeff Jarvis Download and subscribe to AI Inside in audio and video: https://aiinside.show/ Support the podcast on Patreon for special perks: https://www.patreon.com/aiinsideshow. You'll get ad-free episodes, members-only Discord, T-shirts and stickers you love, and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Learn more about your ad choices. Visit megaphone.fm/adchoices