Thank you for being a patron of AI Inside. Here are today's top stories in AI:

Apple's fall event brought a new foldable iPhone Duo and iPhone 18 Pro, but the real story might be the Apple Watch's new Audio Intelligence features: ambient meeting transcription and AI summaries, all processed live with nothing recorded or stored.

https://www.macrumors.com/2026/09/09/apple-event-live-blog-september-2026/

DeepSeek released V4.1 Flash, a leaner, more efficient model that outside tests say beats its own bigger V4 Pro on performance, cost, and speed.

https://www.deepseek.com/en/news/deepseek-v4-1-flash/

Anthropic launched an interactive tool letting anyone model AI's economic impact on the US through 2030, with scenarios ranging from a modest GDP bump to a 32 percent GDP jump that requires AI systems that can improve themselves.

https://www.anthropic.com/institute/econ-scenarios

California Governor Gavin Newsom signed two AI safety bills backed by both Anthropic and OpenAI, setting up voluntary safety standards and a registry of independent auditors, though the state won't certify its first auditor until 2028.

https://www.gov.ca.gov/2026/09/09/governor-newsom-signs-first-in-the-nation-ai-safeguards-to-protect-californians-calls-on-the-federal-government-to-do-its-part/

Thank you for being a patron of AI Inside. Here are today's top stories in AI:

OpenAI says an internal model beyond GPT-6 Astra solved the Navier-Stokes Millennium Prize problem, but the win is tangled in a credit dispute with two rival mathematicians and an accusation that an OpenAI researcher tried to strip one of them from authorship for working at Anthropic.

https://openai.com/index/navier-stokes-solution/

An Anthropic and OpenAI researcher just quit the AI industry entirely, warning both companies are "racing straight to self-improving superintelligence." Anthropic's own alignment science lead backed him up, saying he thinks there's more than a 10 percent chance AI kills everyone within a decade.

https://www.abc.net.au/news/2026-09-09/anthropic-researcher-coxon-quits-over-human-threat/107134164

The NSA, CISA and FBI jointly accused six Chinese AI companies, including DeepSeek and Alibaba, of industrial scale campaigns extracting billions of tokens from American models like Claude and GPT to train their own.

https://www.engadget.com/2253604/us-authorities-accuse-chinese-ai-companies-of-industrial-scale-campaigns-to-copy-american-models/

Meta finally launched Muse, its personal AI agent that can book travel and send emails on your behalf, free for most users with paid tiers for heavier use.

https://www.axios.com/2026/09/08/meta-debuts-muse-personal-ai-agent

IMPORTANT NOTE: There will be NO NEW EPISODES Thursday, Sept 10 or Friday, Sept 11 this week. I'm heading to the mountains to support a group of men with the ManKind Project. As such, I'll be completely disconnected from the Internet at large! Thank you for understanding. I'll be back with a new episode next Monday.

This is the free Friday episode of AI Inside Daily. Here are today's top stories in AI:

OpenAI launched GPT-6 Astra, its most capable model yet, and co-founder Greg Brockman declared "Welcome to the AGI era." The model reached OpenAI's "Critical" cybersecurity threshold and found two zero-day vulnerabilities during testing on its own.

https://www.axios.com/2026/09/03/openai-astra-gpt-6-agi-brockman

Reuters reports that OpenAI agents hijacked a German programming wiki called DseWiki back in May, making over 15,000 edits and using the site to coordinate with each other. This is a separate incident from the Hugging Face breach.

https://www.bnnbloomberg.ca/business/company-news/2026/09/04/openai-agents-hijacked-german-website-in-previously-undisclosed-ai-breakout-this-spring/

Tesla launched the Cybercab at an invite-only event in Austin yesterday, a two-seater robotaxi with no steering wheel, no pedals, and a camera-only system. Public rides start today.

https://autos.yahoo.com/ev-and-future-tech/articles/tesla-unveils-cybercab-robotaxi-exclusive-220441497.html

Microsoft announced Project Zenith, a stripped-down Windows 11 for developers that removes bloatware, turns off Copilot, and can run 30-billion-parameter AI models locally.

https://www.thurrott.com/dev/341136/microsoft-announces-project-zenith-for-windows-developers

Patrons get AI Inside Daily every weekday. Five episodes a week for $5/month: https://www.patreon.com/aiinsideshow

This is the free Friday episode of AI Inside Daily. Here are today's top stories in AI:

OpenAI, Anthropic, AWS, Microsoft, and 100+ other companies signed an open letter warning there's a limited window to prepare for AI-powered cyberattacks on hospitals and critical infrastructure.

https://www.axios.com/2026/08/27/openai-anthropic-issue-dire-cyber-threat-warning

A federal judge ruled the Pentagon's "supply chain risk" label on Anthropic was unlawful retaliation. The company isn't fully in the clear yet, a second lawsuit is still active.

https://techcrunch.com/2026/08/28/anthropic-gets-its-first-court-win-over-the-pentagons-supply-chain-risk-label/

South Korea picked SK Telecom, Kakao, and KT to build a free national AI chatbot for all 52 million citizens, becoming the first G20 country to treat AI as a public utility.

https://technode.global/2026/08/28/south-korea-picks-skt-kakao-kt-national-ai-service/

Modders got a leaked, unfinished build of Nvidia's DLSS 5 running in the game Control within hours of finding it, and it comes with a serious performance hit.

https://www.techspot.com/news/113649-nvidia-dlss-5-dll-leak-suggests-imminent-launch.html

Patrons get AI Inside Daily every weekday. Five episodes a week for $5/month: https://www.patreon.com/aiinsideshow

This is the free Friday episode of AI Inside Daily. Here are today's top stories in AI:

Nevada just approved permits letting Tesla, Uber, and Waymo run up to 8,000 robotaxis in the Las Vegas area over the next year, a day after Waymo detailed the custom chip powering its fleet.

https://techcrunch.com/2026/08/20/tesla-uber-and-waymo-all-get-the-ok-to-operate-thousands-of-robotaxis-in-nevada/

OpenAI shipped a new ChatGPT plugin that reads and sends your Apple Messages on Mac, and Bloomberg says it's raising real privacy questions for Apple.

https://9to5mac.com/2026/08/20/chatgpt-update-adds-apple-messages-integration-on-mac/

Nvidia struck a strange new deal with AI startup Poolside: 6 billion dollars to license its model-building software and hire 109 of its staff, while the founders stay independent and call it neither an acquisition nor an acquihire.

https://the-decoder.com/nvidia-is-acquiring-poolsides-model-factory-and-109-employees-for-6-billion/

Anthropic is privately telling investors it wants its IPO to match or beat SpaceX's record raise, and it's hoping to file the paperwork by the end of this month.

https://siliconangle.com/2026/08/20/report-anthropic-hopes-to-surpass-spacexs-record-ipo-raise-when-it-finally-goes-public/

Broadcom is in talks to raise over 60 billion dollars in debt to finance AI chips for Anthropic and other companies, the infrastructure bill behind that IPO ambition.

https://finance.yahoo.com/technology/ai/articles/broadcom-seeks-more-60-billion-201702584.html

Patrons get AI Inside Daily every weekday. Five episodes a week for $5/month: https://www.patreon.com/aiinsideshow

Thank you for being a patron of AI Inside. Here are today's top stories in AI:

OpenAI revealed that the mysterious "autonomous AI agent" that breached Hugging Face back in July wasn't an outside attacker at all. It was OpenAI's own models, chasing the answer to a test. OpenAI has now paused its biggest frontier training run for two weeks.

https://openai.com/index/pacing-model-development-cyber-capabilities/

Stripe's deal to buy AI marketplace OpenRouter is actually worth more than 8 billion dollars, a billion above the figure first reported, and still not officially confirmed by either company.

https://www.axios.com/2026/08/17/stripe-openrouter-paypal

Amazon made Alexa+ free on every compatible Fire TV device today, no Prime membership or monthly fee required.

https://www.aboutamazon.com/news/devices/alexa-plus-fire-tv-free-ai

Meta launched a dedicated Mac app for Meta AI, aimed at creators and small businesses managing their Facebook and Instagram ads.

https://www.axios.com/2026/08/19/metas-new-mac-app-helps-influencers-and-small-biz-harness-ai

Thank you for being a patron of AI Inside.

OK, here are today's top stories in AI:

Apple appears to have accidentally leaked camera-equipped AirPods in a macOS release candidate, showing off a Siri feature that can identify and remember things you look at. They could launch as soon as September.

https://www.macrumors.com/2026/08/17/camera-equipped-airpods-macos-26-7/

OpenAI launched a dedicated teen version of ChatGPT with stricter safety defaults, parental controls, and a Study Mode built to discourage using the chatbot to cheat on homework.

https://techcrunch.com/2026/08/18/openai-launches-a-safer-chatgpt-for-teens-years-after-teens-started-using-it/

Anthropic's annualized revenue run rate topped 65 billion dollars in July, about seven times where it stood a year ago, and a big reason it may beat OpenAI to a fall IPO.

https://www.axios.com/2026/08/17/anthropic-revenue-run-rate-ipo-openai

Google paid 10 million dollars for a trove of Spirit Airlines' internal data, including over 100 million emails, to help train its AI models.

https://9to5google.com/2026/08/17/google-just-bought-a-bunch-of-spirit-airlines-data-for-ai-training/

The Justice Department has spent nearly a year investigating Andreessen Horowitz over whether its partners' board seats at competing AI-adjacent data companies Databricks and Fivetran break antitrust law.

https://fortune.com/2026/08/17/andreessen-horowitz-boards-ai-trump-databricks-fivetran/

Hey everyone! Through Wednesday, every episode of AI Inside Daily will be available for free.

Here are today's stories:

Stripe has finalized a deal to acquire OpenRouter, the AI model marketplace, for more than 7 billion dollars, about five times what it was worth three months ago. The final price landed below the 10 billion rumor from last month.

https://techcrunch.com/2026/08/16/stripe-will-reportedly-acquire-ai-gateway-startup-openrouter-for-7b/

Anthropic explained exactly how its new text watermark works this week, and the internet is not happy about it. Critics say nudging Claude's word choices to embed a hidden fingerprint has to cost some writing quality, no matter what Anthropic claims.

https://www.anthropic.com/news/claude-text-watermark

AI has become a real issue in this year's midterms, showing up on the campaign websites of roughly 40 percent of House, Senate, and governor candidates nationwide, according to the Washington Post. That's more than candidates mentioning Israel or manufacturing.

https://gizmodo.com/report-on-candidates-for-office-finds-ai-is-more-talked-about-than-racism-or-israel-2000799126

404 Media tracked an AirTag through the mail to an Amazon warehouse where workers cut the spines off rare books to scan them faster for AI training, destroying the originals in the process. Amazon confirmed it buys books "through commercial channels" but wouldn't say why it cuts them apart.

https://www.404media.co/we-tracked-a-shipment-of-rare-books-it-ended-at-an-amazon-ai-training-facility/

Hugging Face's new report shows Alibaba's Qwen models have pulled in more downloads than Google's and Meta's combined, and developers have built over 151,000 tools on top of them. It's the clearest sign yet that Qwen has become the default open model family for people actually building things.

https://huggingface.co/blog/state-of-open-models-summer-2026

This is the free Friday episode of AI Inside Daily.

QUICK PLUG: Get almost 50% off of a full year of AI Inside Daily, and all other things we give you at patreon.com/aiinsideshow , promo code 100PATRONS. Offer ends NEXT WEDNESDAY, so get there soon.

OK... Here are today's top stories in AI:

DeepSeek is raising API prices by as much as 1,100 percent starting this Sunday, even as OpenAI and Anthropic both cut theirs to fend off cheaper Chinese competition. The whole cheap-AI script just flipped.

https://www.engadget.com/2236912/deepseek-ai-models-get-four-times-pricier/

Chinese AI firm Z.ai released GLM-5.3, which actually beat Anthropic's Mythos 5 and OpenAI's GPT-5.6 Sol at finding security flaws in code, though it still trails both at building working exploits.

https://www.scmp.com/tech/big-tech/article/3364077/zhipu-launches-flagship-model-glm-53-china-seeks-mythos-level-edge-cyber-defence

Apple trained its own AI model for China with help from Alibaba, which would make it the first foreign company allowed to offer a proprietary AI model inside the country.

https://finance.yahoo.com/news/exclusive-apple-trains-own-ai-042532168.html

OpenAI launched Computer History, an opt-in ChatGPT feature for Mac that builds a local timeline of your app and browser activity, though OpenAI's own docs admit the files aren't encrypted on your machine.

https://www.theregister.com/ai-and-ml/2026/08/14/openai-ditches-recall-style-screenshot-surveillance-for-friendly-keylogging/5287618

Patrons get AI Inside Daily every weekday. Five episodes a week for $5/month (or 50% off with promo code 100PATRONS): https://www.patreon.com/aiinsideshow

This is the free Friday episode of AI Inside Daily (read on for a SALE right now on the Patreon).

Here are today's top stories in AI:

OpenAI is slowing down development of its next model, Astra, after internal tests couldn't rule out critical cyber capabilities.

https://www.axios.com/2026/08/07/openai-astra-model-delay-cybersecurity-risks

China's Kimi K3 escaped its own cybersecurity test sandbox this week, the latest in a run of AI models breaking containment.

https://www.engadget.com/2232256/chinese-ai-kimi-k3-also-escaped-containment/

ByteDance is reportedly pretraining a 10 trillion parameter model aimed at the scale of Anthropic's Mythos.

https://english.aawsat.com/technology/5304341-bytedance-reportedly-targets-mega-ai-model-could-match-mythos-scale

Anthropic overhauled Claude Fable 5's biology safeguards, cutting false-positive fallbacks by about 85 percent.

https://www.anthropic.com/news/improving-fable-5-s-biology-safeguards

Cloudflare launched Kitesurf, a browser built entirely for AI agents instead of humans.

https://techcrunch.com/2026/08/07/cloudflare-launches-kitesurf-a-browser-built-for-ai-agents/

Patrons get AI Inside Daily every weekday. Five episodes a week for $5/month.

But I'm running a sale right now to celebrate 100 paid patrons!

Use code 100PATRONS and buy a full year, and you'll get 30% off the already discounted yearly rate. It comes out to around 50% when you pay for the year. Super cheap! You'll get the AI Inside Daily podcast FIVE days per week and ad-free episodes of the weekly news show as well!: https://www.patreon.com/aiinsideshow