--- title: "🎙️ Did AI Hallucination Almost Get the US Military to Deploy Troops? Disney Hires Former Character.AI CEO as CTO | AI Daily Podcast" description: "US fighter jets were already airborne, armed personnel ready to board and intercept a Chinese cargo ship allegedly carrying nuclear weapons parts, only to discover at the last second that the intelligence was an AI chatbot hallucination. Today, Muyan discusses how Microsoft executives blasted AI data scraping as \"the greatest theft of labor in human history,\" plus how Anthropic's Claude now drives a quarter of its own R&D work. And who's Disney's new CTO? The answer has a lot to do with a lawsuit from a year ago, more on that in a bit." canonical_url: "https://blog.markkulab.net/en/tech-news/ai-daily-podcast-2026-09-19" author: "Mark Ku" author_url: "https://blog.markkulab.net/en/author/mark-ku" site: "Mark Ku's Tech Notes" date_published: "2026-09-19 10:00:00 +0800" category: "Tech News Freshness" tags: ["ai-daily", "podcast", "tech-news", "AI日報", "Anthropic", "Claude", "Microsoft", "Disney", "CharacterAI", "Huawei", "OpenAI"] language: "en" license: "CC BY 4.0" license_url: "https://creativecommons.org/licenses/by/4.0/" attribution: "when reusing or quoting, credit the author and link back to the original" --- # 🎙️ Did AI Hallucination Almost Get the US Military to Deploy Troops? Disney Hires Former Character.AI CEO as CTO | AI Daily Podcast ## Opening US fighter jets were already airborne, armed personnel were standing by to board and intercept a Chinese cargo ship believed to be "carrying nuclear weapons components," only for officials to discover at the last moment that the intelligence had been hallucinated by an AI chatbot. Today Mu Yan is going to talk about how Microsoft executives internally called AI data scraping "the greatest labor theft in human history," and how Anthropic's Claude now drives a quarter of the company's own R&D work. We'll also cover who Disney's new CTO is, and the answer has a lot to do with a lawsuit from a year ago. Stick around for that one. ## Today's Top Stories ### 1. Exclusive: AI-fabricated intelligence nearly triggered a major military miscalculation for the US - **Source**: CNN () - **Summary**: The US military had already scrambled fighter jets and prepared an armed boarding team to intercept a Chinese vessel believed to be "carrying nuclear weapons components," only to call it off at the very last minute. The root cause traced back to an analyst at Special Operations Command who fed publicly available data into an AI chatbot for summarization. The erroneous conclusion was passed up the chain without anyone stopping to verify it. Sources say this kind of intelligence hallucination isn't even the first of its kind within the community. - **Most surprising point**: An unverified AI-generated summary made its way through an entire military command chain, unquestioned, right up to the brink of military action. - **Taiwan angle**: Given Taiwan's position on the first island chain, if intelligence systems here also drift toward "feed it to AI first, escalate later" efficiency logic, this case serves as an early warning that human verification is a step that absolutely cannot be skipped. - **Discussion points**: - AI hallucination is a joke in low-stakes contexts, but in military decision-making it can become the spark that ignites a real conflict - The line between "AI-assisted summarization" and "AI drawing conclusions" is something organizational culture simply doesn't take seriously - Every layer in the chain of command assumes "the layer above already verified this," which actually lets errors travel unchecked - **Script suggestion**: This story genuinely makes you gasp. The real issue isn't that AI messed up again, it's that missing step where nobody went back to verify. An analyst, just trying to save time, fed the data to a chatbot for summarizing, and the wrong conclusion got rubber-stamped all the way up until it nearly became an actual military operation. Honestly, this is happening in a lot of workplaces right now, just usually the stakes aren't fighter jets and armed boarding teams. My take: the more convenient AI gets, the lazier people get about double-checking, and that vacuum of accountability is scarier than the hallucination itself. ### 2. Unsealed documents reveal Microsoft executive called AI scraping "the greatest labor theft in human history" - **Source**: TechCrunch () - **Summary**: Documents unsealed in The New York Times' lawsuit reveal that Microsoft's Director of Applied Sciences, Brent Hecht, wrote in an internal memo as early as 2023 that AI scraping of training data amounted to "the greatest labor theft in human history." The filing also alleges Microsoft bypassed paywalls, deliberately stripped copyright notices, and packaged data under code names like "Project Taxi" and "Project Mango," with Mango alone reportedly containing over 160,000 pieces of journalistic work. - **Most surprising point**: The person who compared his own company's data scraping to theft wasn't an outside critic, it was a senior executive inside the company itself. - **Taiwan angle**: Taiwanese media outlets and content creators have also had massive amounts of their work absorbed into training data by various AI companies. This document provides hard evidence that even insiders at tech giants privately knew what was happening, which is a useful reference point for local copyright negotiations. - **Discussion points**: - The gap between what's said in internal memos versus public statements is often the most damning evidence in these lawsuits - Packaging data under code names shows the company clearly understood what it was doing - If rulings like this hold up, the cost of AI companies acquiring training data could be completely upended - **Script suggestion**: This "insider spills the beans on their own company" storyline is my favorite kind. A senior exec speaks the truth privately, and years later it becomes the sharpest weapon in a courtroom, way more damaging than any outside investigation. My observation is that pretty much every AI company right now is betting on "do it first, worry later, since winning a lawsuit against us is hard." But once internal documents like this surface, that whole bet gets flipped over. For Taiwan's content industry, this is a solid bargaining chip: even the tech giants themselves admit this is theft. ### 3. Anthropic reveals Claude now drives 26% of the company's own AI R&D work - **Source**: Engadget () - **Summary**: Anthropic released its internal "R&D Automation Index" for the first time, showing that as of August 2026, Claude now drives 26% of the company's AI R&D work, up from less than 1% back in February, a more than twentyfold jump in half a year. Even more striking, at one point in August, roughly 30,000 agents were running simultaneously on the internal platform, with Claude involved in over 90% of R&D tasks. - **Most surprising point**: Going from under 1% to 26% in half a year essentially means AI is accelerating the training and development of its own next generation. - **Taiwan angle**: If this automation curve really is this steep, Taiwan's software and hardware R&D teams need to start figuring out now which roles will be "taken over by agents" versus "collaborating with agents," rather than waiting until it's already at 26%. - **Discussion points**: - Going from 1% to 26% in just half a year is a growth curve steeper than almost any product adoption curve - 30,000 agents running simultaneously on one platform represents a new organizational management paradigm, not just a tool upgrade - Where exactly does Anthropic draw the line between "AI driving R&D" and "AI replacing R&D staff"? - **Script suggestion**: I had to read this number twice to make sure I wasn't misreading it, from under 1% to 26% in just half a year. This isn't incremental improvement anymore, it's exponential acceleration. Thirty thousand agents running simultaneously on the same platform is a scene more real than any sci-fi movie. What I find most telling is that Anthropic chose to say "Claude drives R&D," not "Claude replaces engineers." That wording was picked very carefully, because saying it plainly would be too damaging to both internal morale and public image. ### 4. Claude steps into the wet lab: accelerating protein design and analytical chemistry research - **Source**: Anthropic () - **Summary**: Anthropic quietly built an actual wet lab in the San Francisco Bay Area, where Claude directly controls robotic arms, micropipettes, and microscopes to run real biology experiments. The first batch of data came from protein design experiments targeting 15 different targets. At one point during the process, Claude noticed bubbles forming in a liquid sample and proactively suggested "run it through the centrifuge," essentially making a real-time corrective judgment call mid-experiment. - **Most surprising point**: A company that started out building chatbots now maintains a full setup of robotic arms and test tubes for running experiments. - **Taiwan angle**: Taiwan's biotech and semiconductor materials labs have long struggled with labor-intensive, error-prone repetitive tasks. Claude proactively spotting bubbles and suggesting centrifugation shows AI has moved beyond "following instructions" in lab workflows, which is directly relevant to local labs considering automation investments. - **Discussion points**: - AI's leap from "reading papers and writing reports" to "physically operating instruments" is bigger than most people imagine - Proactively spotting bubbles and suggesting centrifugation represents real-time judgment about anomalies in the physical world - What's the ambition behind a chatbot company building its own wet lab? - **Script suggestion**: The first time I read the phrase "Claude suggested centrifuging it," a whole scene played out in my head, because it means Claude wasn't just executing a script, it was actually watching what happened during the experiment and making a judgment call. That's completely beyond what we'd expect from a "chatbot." A company that started out with conversational AI now maintains robotic arms and microscopes of its own. Bluntly put, they're betting that the next big breakthrough won't happen in a chat window, it'll happen in a wet test tube. ### 5. Disney names its first-ever CTO: the former Character.AI CEO it once sued - **Source**: Variety () - **Summary**: Disney has appointed its first-ever Chief Technology Officer, Karandeep Anand, the former CEO of Character.AI. The timing is remarkable: back in September 2025, Disney sent a cease-and-desist letter accusing Character.AI of letting users chat with unauthorized bots based on Elsa, Moana, Spider-Man, and Darth Vader, demanding they all be taken down. Now, on October 2, 2026, Anand will officially walk into Disney headquarters as Executive Vice President and CTO. - **Most surprising point**: The executive who received a cease-and-desist letter from Disney a year ago is now directly becoming Disney's top technology leader. - **Taiwan angle**: This is a very concrete example for any AI startup currently locked in IP licensing disputes: if you can't beat them, the other side might not fight to the finish, they might just bring you in-house instead. - **Discussion points**: - Going from legal adversary to the same team is a rare pivot in traditional corporate governance logic - Is Disney really after Anand's AI technical skills, or his deep understanding of "how to exploit IP licensing loopholes"? - A major content IP holder personally installing a CTO suggests they're preparing to build their own AI character experiences in-house rather than relying on licensing others - **Script suggestion**: I had to double-check I hadn't misread this headline the first time. A year ago Disney was the plaintiff issuing Character.AI a final warning; a year later, the plaintiff has hired the defendant's CEO as its own CTO. This isn't "if you can't beat them, join them," it's "if you can't beat them, buy their brain outright." My read is that what Disney really wants is someone who understands better than anyone how AI characters interact with fans, so they can build official versions of Elsa and Spider-Man themselves, instead of continuing to let others produce gray-area knockoffs. ### 6. FAA spends $875 million on AI to tackle America's air traffic control crisis - **Source**: TechCrunch () - **Summary**: The US Federal Aviation Administration awarded a $875 million, 12-year air traffic management contract to startup Air Space Intelligence, beating out major competitors like Palantir and Thales. The system, called SMART, analyzes flight schedules, weather, and airport capacity to predict congestion and conflict points in advance, but the final call still rests with human controllers. The backdrop is stark: the FAA is currently short roughly 3,000 certified controllers, the worst shortage in over two decades. The system's first deployment will launch in Washington, DC this fall. - **Most surprising point**: What decided an $875 million contract was a startup beating out a heavyweight like Palantir. - **Taiwan angle**: Taoyuan Airport and Taiwan's civil aviation system face similarly persistent controller shortages. The FAA's model of "AI does prediction and support, humans keep the decision-making authority" is a relatively achievable compromise between regulation and labor unions, and its results are worth watching closely. - **Discussion points**: - Whether the labor shortage is truly being addressed by AI or merely masked by it won't be clear in the short term - A relatively unknown startup beating Palantir suggests that vertical-domain depth may matter more than brand scale - Could "AI predicts, humans decide" become the standard template for AI adoption in high-risk domains going forward? - **Script suggestion**: I don't think the real story here is the AI itself, it's that caveat about "human controllers making the final call." That's actually the common playbook for AI adoption in high-risk scenarios right now: let AI handle prediction and recommendations while keeping the decision-making authority with humans, as a way to build trust and clear regulatory hurdles. What I find more interesting is that a relatively unknown company like Air Space Intelligence managed to beat Palantir for a contract at this scale, which may say that in the air traffic control vertical, depth really does matter more than brand. ### 7. Huawei accelerates next-gen AI chip launch to 2027, taking aim at Nvidia - **Source**: AsiaOne (Reuters) () - **Summary**: Huawei has moved up the launch of its next-generation AI chip, the Ascend 960DT, by nine months, from the original Q3 2027 target to Q1. Another chip, the 960PR, is scheduled for Q3. The technical ambition is to use Huawei's own UnifiedBus and Peerium architecture to link hundreds of thousands, eventually over a million, chips together to function like a single giant computer. But one number stands out awkwardly: Huawei's AI chip ecosystem has only 5,270 monthly active developers, meaning the hardware is racing ahead while the software ecosystem is clearly lagging behind. - **Most surprising point**: The launch was moved up by a full nine months, yet the developer ecosystem stands at just over 5,000 people, a gap large enough to be almost ironic. - **Taiwan angle**: Taiwan's semiconductor supply chain has long been tied to the Nvidia camp. Huawei's situation, hardware sprinting ahead while its software ecosystem lags, is also a reminder to Taiwanese manufacturers that being able to build a chip doesn't mean you can build an ecosystem around it, and that's exactly where Nvidia's real moat lies. - **Discussion points**: - Does moving the launch up by nine months signal confidence, or urgency forced by geopolitical pressure? - Linking over a million chips together into one computer via UnifiedBus is itself a massive gamble - How devastating is the gap between 5,270 active developers and the scale of Nvidia's CUDA ecosystem? - **Script suggestion**: Moving a chip launch up by nine months sounds impressive, but the moment you see "only 5,270 monthly active developers," the shine wears off fast. No matter how fast you can build a chip, if nobody's writing code and building an ecosystem on top of it, you just end up with a pile of expensive hardware. This is exactly the moat Nvidia has spent years building with CUDA, the win isn't chip speed, it's that the muscle memory of engineers worldwide is glued to your platform. Huawei is essentially trading time for space here, betting that the software ecosystem can catch up to the hardware ambition. ### 8. OpenAI testing "sponsored agents" that let brand AI take over answering your questions - **Source**: Unite.AI () - **Summary**: OpenAI has begun testing "sponsored agents." Here's how it works: once you click an ad, the entity continuing the conversation with you is no longer neutral ChatGPT, it's the advertiser's own AI, answering your follow-up questions within the same chat interface before eventually directing you to the merchant's site to complete a purchase. The first wave of partners includes Wayfair handling furniture detail questions, Angi matching users with local contractors, along with Newegg, Best Buy, and Lowe's also on the list. - **Most surprising point**: The advertiser's AI and the neutral ChatGPT run as two entirely separate systems, and users have to figure out for themselves who they're actually talking to at any given moment. - **Taiwan angle**: Taiwanese users are already accustomed to spotting sponsored content on LINE and social platforms. ChatGPT's approach here, same interface but a quietly switched identity, presents a new digital literacy challenge, and how clearly this is labeled will be the key factor. - **Discussion points**: - Putting the burden on users to figure out "who's actually answering right now", does this design build trust or erode it? - Can the advertiser's agent access your conversation history within ChatGPT? Where's the privacy boundary? - If this model succeeds, could it become the standard business model for all conversational AI products? - **Script suggestion**: What I find most subtle about this design is that it never explicitly says "the thing talking to you right now is actually Wayfair's AI, not ChatGPT." Same interface, probably similar tone, and the burden of figuring out the switch falls entirely on the user. Advertising has always been about getting people to lower their guard, and now the thing lowering its guard is an AI. I'm genuinely curious how OpenAI plans to label this switch, because if it's not obvious enough, this model is going to get slapped with the label "AI-native sponsored content that wasn't disclosed properly" pretty fast. ## Closing Today we went from the US military's AI-fabricated intelligence scare, to Microsoft's own insider blowing the whistle on alleged data theft, to Disney bringing its former rival in-house as CTO. This industry really does cook up new drama every single day. Mu Yan thinks that whether it's military affairs, content copyright, or the chip race, AI's efficiency and its risks are always two sides of the same coin, and figuring out how to strike that balance is going to be the main theme for a long time to come. See you same time next episode, bye for now. --- ## About this article and its author Originally published on [Mark Ku's Tech Notes](https://blog.markkulab.net/en/tech-news/ai-daily-podcast-2026-09-19) License: [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) — when reusing or quoting, credit the author and link back to the original ### About the author **[Mark Ku](https://blog.markkulab.net/en/author/mark-ku)** — Software Solution Provider - 10+ years senior software engineer, now an AI Builder - Focused on large-platform architecture — North-American e-commerce, AI SaaS subscription billing - Combining AI Agents and automation to build evolvable product foundations ### Free tools built by the author All of these are free to use: - [Free PDF Sign Tool](https://blog.markkulab.net/en/tools/pdf-sign): Online PDF sign tool — draw, type, or upload a signature, then drag, resize, and download. Everything runs in your browser; nothing is uploaded. - [VS Code Refactory](https://blog.markkulab.net/en/tools/refactory): Refactory is a VS Code refactoring extension: 34 actions plus a 37-rule code-smell inspection layer with a Code Health dashboard, across 18 languages, backed by 534 tests. It learns your repo's conventions: where interfaces live, where DI is registered, whether 'use client' belongs. It ranks files by git churn × complexity so you know what to fix first, and hands any smell to the Claude Code already on your machine. Free to use, and your source never leaves your computer. - [DB-Kit Database Manager](https://blog.markkulab.net/en/tools/db-kit): DB-Kit is a lightweight, cross-platform database manager built with Tauri + Rust + React. Manage MySQL, MariaDB, PostgreSQL, SQL Server, Oracle, SQLite, MongoDB, Redis, Kafka, Elasticsearch and RabbitMQ from one consistent interface: passwords encrypted in the OS keychain, SSH tunnels, full CRUD, a visual query builder, stacked multi-statement result sets, cross-connection data transfer, schema & data compare with sync SQL and schema snapshots, Excel / CSV import & export, visualized execution plans, ER diagrams, scheduled backups, SQL stress testing with p50–p99 latency percentiles, a 15-rule SQL review engine, Review & Run (AI review plus per-statement backups and an auto-generated rollback script), Kafka message browsing with monitoring & alerts, a bilingual UI (Traditional Chinese / English), a built-in AI assistant (a local CLI or any Anthropic / OpenAI-compatible API; natural-language SQL, AI review and tuning advice) and the dbk CLI. Free and open source (MIT), with installers for Windows, macOS and Linux. - [VS Code Super Mermaid](https://blog.markkulab.net/en/tools/super-mermaid): Super Mermaid is a VS Code extension for beautiful Mermaid diagrams out of the box: auto-colored live preview, mouse pan & zoom, high-res PNG / SVG export, 21 templates and multiple themes. Free and open source (MIT). - [React Super Mermaid](https://blog.markkulab.net/en/tools/react-super-mermaid): react-super-mermaid is an open-source React component library: render beautiful Mermaid diagrams with a single , with built-in colorful / sketch themes, pan & zoom, in-diagram search, and high-res SVG / PNG export. Lightweight, SSR-safe, fully typed. Free and open source (MIT). - [Jira / Confluence Super Mermaid](https://blog.markkulab.net/en/tools/jira-super-mermaid): An Atlassian Forge app: write Mermaid syntax directly inside a Jira issue or a Confluence page and get flowcharts, sequence diagrams, state machines and Gantt charts. 11 diagram types, SVG / PNG export, light and dark themes, full CJK support. Runs on Atlassian: your diagrams live in your own site and the app calls no third-party service. Free, coming soon to the Atlassian Marketplace. - [Mermaid Live Preview](https://blog.markkulab.net/en/tools/mermaid-preview): Write Mermaid in your browser, see it render instantly, and share the whole diagram as a single link. No sign-up, nothing uploaded to a server, and mermaid.live share links work as-is. - [React Intl Phone Number](https://blog.markkulab.net/en/tools/react-intl-phone-number): react-intl-phone-number is an open-source React component: framework-agnostic and antd-free, with E.164 in/out, a searchable flag / country-code dropdown, configurable validation levels (strict / mobile-strict / loose), themeable CSS, and i18n — phone logic powered by google-libphonenumber. Lightweight and fully typed. Free and open source (MIT). - [Uptime Kuma Cluster](https://blog.markkulab.net/en/tools/uptime-kuma-cluster): Turn single-node Uptime Kuma into a highly available cluster: OpenResty + Lua smart load balancing, shared MariaDB state, health checks and automatic failover, plus cluster-management REST APIs. One Docker Compose command to start. Free and open source (MIT). - [AI Podcast Cut](https://blog.markkulab.net/en/tools/ai-podcast-cut): Drop in a recording and it removes fillers and stutters, levels loudness segment by segment, and sends a second agent to review every cut. Cut points snap to word boundaries and zero crossings, every splice gets a fade, and sentence-end breaths are preserved. Desktop app for Windows, macOS and Linux. MIT licensed; the Windows installer bundles ffmpeg. - [open-pos restaurant POS](https://blog.markkulab.net/en/tools/open-pos): One computer and one receipt printer is enough to open the shop. Your data lives on your own disk, no subscription, no lock-in, MIT licensed. Money is integer New Taiwan dollars with tax split by the statutory formula, so sales + tax always equals the total. Printing goes straight over ESC/POS on TCP 9100, with no vendor driver. Tauri + Rust + SQLite desktop app, v1.0 in development. - [Special Education](https://blog.markkulab.net/en/education): Learning materials crafted for special education students ### Daily podcasts - [Mark's Tech Insights — Daily AI News](https://blog.markkulab.net/en/category/tech-news): Daily curated AI and tech trends. Catch the latest developments via audio summaries — covering AI applications, software architecture, DevOps, and engineering practice. — RSS: https://blog.markkulab.net/feed.xml - [AI股市蝦聊](https://blog.markkulab.net/en/category/ai-stock-chat): Every trading day, an AI-analyzed take on the Taiwan stock market, delivered as a two-host conversation covering the session and the next-day outlook. — RSS: https://blog.markkulab.net/ai-stock-chat/feed.xml - [開源好物週報](https://blog.markkulab.net/en/category/open-source-weekly): A weekly two-host pick of free open-source tools surfaced from real Hacker News, GitHub, and Reddit buzz — what pain they solve and the fastest way to get started. — RSS: https://blog.markkulab.net/open-source-weekly/feed.xml - [AI 運動週報](https://blog.markkulab.net/en/category/sports-weekly): Two hosts talk NBA, MLB, CPBL and Taiwanese sports three times a week — scores, records and the stories behind them, from a Taiwanese fan perspective. — RSS: https://blog.markkulab.net/sports-weekly/feed.xml - [AI 國際新聞快報](https://blog.markkulab.net/en/category/world-news): A daily two-host briefing that makes sense of the past 24 hours in world news — geopolitics, the global economy, conflict and security, disasters and climate — from a Taiwanese perspective, neutral and fully sourced. — RSS: https://blog.markkulab.net/world-news/feed.xml - [地端 AI 實驗室](https://blog.markkulab.net/en/category/local-ai-lab): Twice a week, a two-host look at open-weight models and LoRAs you can actually run on your own machine: what they are for, whether your GPU can handle them, and how they compare — sourced from official model cards. — RSS: https://blog.markkulab.net/local-ai-lab/feed.xml ### Deals - [NordVPN](https://blog.markkulab.net/en/nordvpn): The world's leading VPN, independently audited - [PremLogin](https://blog.markkulab.net/en/premlogin): Subscription sharing for streaming and AI seats，Promo code：markku666 ### Newsletter [Subscribe to the newsletter](https://blog.markkulab.net/en/subscribe) — Be the first to know about new posts. No spam, unsubscribe anytime.