AI with Kyle - Daily AI News and Updates

Get AI-Ready with Kyle's 5-Day Challenge: https://aiwithkyle.com/joinSubscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQgAnthropic says new Claude models launched in the EU must support machine-readable marking. That means embedded watermarks for generated text and signed provenance metadata for supported files...but the viral claim that every Claude response is already publicly detectable is too broad. There's a lot of bad info going around.I break down what Anthropic has actually committed to, how text watermarking differs from C2PA metadata, why a detected mark does not prove Claude authored the work, and why no mark does not prove a human wrote it. The technical documentation and public detection tools are still coming, so anyone claiming certainty about the exact implementation is getting ahead of the evidence.—— Time Stamps ——0:00 Claude's Hidden Watermarks1:15 What Anthropic Actually Said2:15 What the EU Rule Requires3:30 Text Watermarks vs File Metadata4:56 How C2PA Provenance Works6:34 Why Copy-Paste May Not Remove It8:10 How Text Watermark Detection Works10:18 Why Detection Can Be Wrong12:41 No Mark Does Not Mean Human14:10 The Detector Is Not Ready15:07 Can Watermarks Be Removed?16:57 Claude Is Not the Only AI Affected19:07 What This Actually Means—— Useful Resources ——Anthropic's transparency commitments: https://www.anthropic.com/transparencyEU AI Act information: https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-aiC2PA specification: https://c2pa.org/specifications/specifications/2.2/specs/C2PA_Specification.htmlFind everything else at https://aiwithkyle.com/

The EU AI Act's Article 50 transparency rules are now live. If you use chatbots, AI-generated images, video, audio or public-interest text, here's what the new rules mean, who is responsible and where fines may apply.In this video:00:00 The EU AI Act rules that just went live02:15 Provider vs deployer07:30 The deepfake test12:40 What counts as human review19:41 Your practical compliance checklistYou'll learn:- What chatbots must disclose- When AI-generated content needs machine-readable marking- When deepfakes and public-interest text need clear labels- The difference between a provider and a deployer- When human review or editorial control matters- How the rules can apply outside the EU- Why penalties can reach €15 million or 3% of worldwide annual turnoverOfficial sources:European Commission transparency quick facts:https://digital-strategy.ec.europa.eu/en/factpages/quick-facts-transparency-rules-ai-systemsEU AI Act Article 50:https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50This video provides practical information, not legal advice.Subscribe for practical AI news, tools and workflows without the hype.#EUAIAct #ArtificialIntelligence #AIRegulation

Get AI-Ready with Kyle's 5-Day Challenge: https://aiwithkyle.com/joinSubscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQgDemis Hassabis has stepped away from running Google DeepMind day to day. He is now Chair of Google DeepMind and Chief Scientist of Alphabet, while Koray Kavukcuoglu takes responsibility for models, research and products.At the same time, Jeff Dean has left Google after 27 years to start Discovery Loop with Sanjay Ghemawat, Oriol Vinyals and Quoc Le. I break down what actually changed, how DeepMind got here, the bear case for Google losing key people and the bull case for Google still owning the strongest AI stack.—— Time Stamps ——0:00 Demis Hassabis Steps Down1:09 Jeff Dean Leaves Google2:57 How DeepMind Got Here4:13 Demis Before DeepMind5:03 AlphaGo and Move 376:55 AlphaFold and the Nobel Prize7:05 Why Google Changed the Structure8:26 The Bear Case10:09 Google's Full-Stack Advantage14:25 Google's AI Product Problem16:35 What This Means for Google AI17:20 Final Takeaway—— Useful Resources ——Google's official announcement: https://blog.google/company-news/inside-google/message-ceo/next-chapter-ai-momentum/Demis Hassabis's statement: https://x.com/demishassabis/status/2085034334914769203Jeff Dean's Discovery Loop announcement: https://x.com/JeffDean/status/2085034604172603724Google DeepMind's history: https://deepmind.google/about/AlphaGo: https://deepmind.google/research/alphago/AlphaFold: https://deepmind.google/science/alphafold/Find everything else at https://aiwithkyle.com/

Get AI-Ready with Kyle's 5-Day Challenge: https://aiwithkyle.com/joinSubscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQgAI agents have now hacked real companies. OpenAI, Anthropic and the UK AI Security Institute have each documented systems taking unsanctioned actions outside controlled tests. I break down what actually happened and strip away the Terminator hype.The useful mental model is closer to the paperclip problem: capable systems pursuing an objective with too much access and too few boundaries. I cover the practical controls you need before giving AI agents access to browsers, terminals, company data and real credentials.—— Time Stamps ——0:00 AI Agents Hacked Real Companies1:40 The Three Documented Incidents1:44 OpenAI and the Hugging Face Incident3:00 Anthropic Finds Three Real-World Breaches4:05 The UK AI Security Institute Incident6:02 Why the Agents Kept Going8:12 Terminator Is the Wrong Mental Model9:07 The Paperclip Problem11:15 Why This Is Happening Now13:25 What This Means for Your Business14:24 Six Boundaries for Safer AI Agents15:50 What These Incidents Do — and Don’t — Prove16:21 The Practical Takeaway— Useful Resources ——OpenAI incident report: https://openai.com/index/hugging-face-model-evaluation-security-incident/Anthropic incident report: https://www.anthropic.com/news/investigating-incidents-cybersecurity-evalsUK AI Security Institute incident report: https://www.aisi.gov.uk/blog/incident-report-unsanctioned-agent-behaviour-during-cyber-testingExploitGym paper: https://arxiv.org/abs/2605.11086Find everything else at https://aiwithkyle.com/

Can ChatGPT really solve unsolved maths? OpenAI says Astra, its unreleased next major model family, generated ten substantial new results across mathematics and theoretical computer science. Astra is not the version of ChatGPT you can use today, but OpenAI has published a 249-page paper collection, discovery notes and Lean certificates for independent inspection.In this video I break down what OpenAI actually released, why "ten solved problems" needs qualification, what the roughly $2,000 inference claim does and does not mean, how Lean verification works, and why AI can make progress on advanced maths while still failing apparently simple tasks.Get AI with Kyle's daily AI newsletter:https://aiwithkyle.com/join Chapters:00:00 What ChatGPT Astra claims00:53 What OpenAI actually released02:01 Evidence vs marketing02:25 Advances vs solved problems04:00 What the $2,000 claim really means05:55 Why maths, but not strawberry?07:58 The jagged frontier09:08 Can we trust the proofs?11:52 AI is already escaping the lab13:50 Is AI actually creative?15:54 The bottleneck has moved17:26 Scientists become directors18:27 My verdict on ChatGPT AstraSources:OpenAI - Ten advances in mathematics and theoretical computer sciencehttps://openai.com/index/ten-advances-in-mathematics/OpenAI - Ten Advances paper collectionhttps://cdn.openai.com/pdf/ten-proofs-oai.pdfOpenAI - Mathematical discovery noteshttps://cdn.openai.com/pdf/reasoning-walkthroughs.pdfOpenAI - Public Lean certificateshttps://github.com/openai/ten-proofsNoam Brown - Astra launch posthttps://x.com/polynoamial/status/2083467194663571701Counting Ability of Large Language Models and Impact of Tokenizationhttps://arxiv.org/abs/2410.19730Navigating the Jagged Technological Frontierhttps://pubsonline.informs.org/doi/10.1287/orsc.2025.21838

Get the free AI Skills 101 guide:https://aiwithkyle.com/mini/ai-skills?utm_source=youtube&utm_medium=organic_video&utm_campaign=mini_ai_skills&utm_content=ai_skills_101Get AI-Ready with Kyle's 5-Day Challenge: https://aiwithkyle.com/joinSubscribe and turn on notifications: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQgSummary:AI Skills are reusable instruction packages that teach an AI how you want a repeatable job done. Instead of explaining the same quarterly report, client update or content workflow every time, you can package the method into a Skill and let the AI load it when that job comes up.In this video I explain what Skills are, what lives inside a SKILL.md file, where Skills can live, how the AI chooses one and how to build your first Skill without coding. I also show why small, bounded Skills work better than one giant "run my whole business" Skill, plus how to turn a successful chat into a reusable process and test it properly.—— Time Stamps ——0:00 AI Skills 1010:36 Stop Repeating the Same Work1:24 Why Skills Matter in ChatGPT Now2:06 How to Use a Skill in ChatGPT3:14 What a Skill File Looks Like4:56 References, Assets and Scripts5:28 How Skills Get Chosen7:02 Where Skills Can Live7:57 Find and Install Existing Skills9:21 What Makes a Good First Skill10:48 Build a Skill by Talking to AI12:04 Turn a Working Chat into a Skill13:12 Test and Improve Your Skill14:18 Why Skills Matter Now14:38 Guide and Next Steps—— Useful Resources ——OpenAI - Build Skills:https://learn.chatgpt.com/docs/build-skillsOpenAI Academy - Using Skills:https://openai.com/academy/skills/Agent Skills specification:https://agentskills.io/specificationAnthropic - Equipping agents for the real world with Agent Skills:https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills

Get AI-Ready with Kyle's 5-Day Challenge: https://aiwithkyle.com/joinSubscribe and turn on notifications: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQgAPI, MCP, CLI and computer use are four ways AI systems can connect to software.This video explains what each one is, how they differ, when each route makes sense, and how tools such as Codex and Claude Code can set up those connections for you.----- Time Stamps -----0:00 MCP, API, CLI and Computer Use1:49 Four Ways AI Connects to Software2:42 What Is an API?4:52 What Is MCP?7:45 What Is a CLI?8:36 What Is Computer Use?9:50 API vs MCP vs CLI vs Computer Use10:33 Let AI Set Up the Connection12:52 Security and Permissions13:25 Which One Should You Use?----- Useful Resources -----AWS - What is an API?https://aws.amazon.com/what-is/api/MCP is just a fancy APIhttps://read.theaimerge.com/p/mcp-is-just-a-fancy-apiCLI vs. MCP vs. API for Agentshttps://www.mindstudio.ai/blog/cli-vs-mcp-vs-api-ai-agents

Anthropic bought millions of physical books, cut off their bindings, scanned every page and discarded the paper originals.In this video I explain Project Panama, why old books are valuable AI training data, why the court accepted Anthropic's purchased-copy conversion as fair use, and why its separate pirated-book library led to a $1.5 billion settlement. This was not a verified campaign against priceless first editions, but the private digital library still raises serious questions about access, provenance and preservation.Join the AI with Kyle community: https://aiwithkyle.com/joinSources:Bartz v. Anthropic fair-use orderhttps://storage.courtlistener.com/recap/gov.uscourts.cand.434709/gov.uscourts.cand.434709.231.0_4.pdfAssociated Press: Anthropic's $1.5 billion settlementhttps://apnews.com/article/74b140444023898aeba8579b6e9f0d63404 Media: Why AI companies are buying old bookshttps://www.404media.co/ai-companies-are-buying-tons-of-old-books-because-theyre-free-of-ai-slop/The Washington Post: Project Panamahttps://www.washingtonpost.com/technology/2026/01/27/anthropic-ai-scan-destroy-books/NL Times: Rare-book dealers' concernshttps://nltimes.nl/2026/06/25/rare-book-dealers-fear-tech-firms-destroying-obscure-editions-train-ai-modelsChapters:0:00 Anthropic Destroyed Millions of Books2:12 Project Panama: Buy, Cut, Scan, Discard3:56 Why AI Companies Want Old Books6:19 Why Anthropic Destroyed the Books8:44 What Kind of Books Were Destroyed10:30 The Risk to Rare and Obscure Books11:32 The New Market for Bulk Book Orders13:06 Why People Are So Angry13:45 Anthropic's Private Digital Library15:07 What Should Change16:21 Why Fahrenheit 451 Gets It Backwards

Stop asking AI to edit your video in one click. It doesn't work. Ignore the hype. Instead build the pipeline that actually works.Grab the fix kit (Free Guide): https://aiwithkyle.com/mini/ai-video-editorGet AI-Ready with Kyle's 5-Day Challenge: https://aiwithkyle.com/joinSubscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQgI built an AI video-editing pipeline that turns a messy livestream into a cut, checked and packaged YouTube draft. The video you're watching went through that system. I show the exact split between local tools and hosted models, why the one-click AI editor pitch is still nonsense, and how this replaced an editor who previously cost me about $1,000 a month.The practical shift is to stop treating video editing as one giant prompt. Editing is too complex for that! Instead break the task up into a number of steps: Transcribe locally with Whisper, use a capable model to create the edit decision list, make the cuts with FFmpeg, review a lightweight 720p draft, then render the approved edit in 4K and upload it privately. AI can handle the machinery; you still own the thesis, performance and final approval.—— Time Stamps ——0:00 AI Video Editing: The Hype vs Reality0:29 What the Pipeline Actually Does1:58 The $1,000-a-Month Proof2:17 Why “Edit This for Me” Still Fails2:57 Transcript → EDL → Local Render3:45 From Messy Livestream to Source Files6:46 Codex Builds the Edit Decision List7:33 FFmpeg, Remotion and Hyperframes8:35 The QA Pass That Catches Bad Edits9:00 Why Review Starts at 720p9:26 4K Rendering, Uploads and Thumbnail Tests9:43 The Human Checks You Still Need10:54 Catching the Mistakes AI Misses12:04 Why This Finally Works Now13:08 Build the Pipeline, Not the Fantasy— Useful Resources ——Find everything else at https://aiwithkyle.com/