A top mathematician has accused OpenAI of privately threatening to ruin his career just to scoop his breakthrough on a major mathematical problem! Welcome to "Mark's Tech Insights." I'm Muyan. Today, besides this explosive tech gossip, we'll also talk about Meta's newly launched AI assistant, Muse. It can read your emails and even swipe your credit card to spend money for you. As for the hidden privacy risks behind it, I'll tell you all about it in a moment.

---

Today's Top Stories

1. OpenAI Accused of "Ugly Behavior"? NYU Mathematician Claims Threats and Theft of Million-Dollar Math Breakthrough

• Source: TechCrunch (https://techcrunch.com/2026/09/08/openai-fought-dirty-on-career-making-math-problem-says-nyu-mathematician/）
• Summary: OpenAI claimed that its internal system of 10,000 parallel agents solved the Navier-Stokes existence and smoothness problem—one of the seven Millennium Prize Problems—in just 88 hours. However, NYU mathematician Tristan Buckmaster alleges that he and a team from Anthropic had already solved it back in August. After privately informing OpenAI, an OpenAI executive pressured him to exclude the co-authors from their competitor, and even threatened him when negotiations broke down, asking: "Why do you want to ruin your career?"
• Most Surprising Part: OpenAI's VP of Research, Sébastien Bubeck, actually said to a mathematician, "Why do you want to ruin your career?"—a threatening line straight out of a mob movie.
• Taiwan Perspective: Taiwan's academic and tech sectors have always highly valued integrity in industry-academia collaboration. This incident serves as a reminder that in the AI era, the intellectual property boundaries of academic achievements are facing immense predatory pressure from tech giants.
• Discussion Points:
  1. Should mathematical proofs generated by AI through brute-force compute hold the same academic standing as the intuitive thinking of human mathematicians?
  2. Now that tech giants monopolize compute power, are they using unfair competitive practices to stifle academic independence?
• Script Suggestion: This is seriously wilder than a Hollywood movie! NYU Professor Tristan Buckmaster alleges that he excitedly called an OpenAI executive to share that he had solved the Navier-Stokes equations. Instead of celebrating, OpenAI not only tried to take all the credit but also demanded he remove his co-authors—who happen to be scientists at Anthropic. The craziest part? When the professor said he would go public, the exec actually threatened him, saying, "Why do you want to ruin your career?" That is straight-up tech mafia behavior! Now, even Fields Medalist Terence Tao can't stand by and watch; he spoke out warning that academia is turning into a cutthroat race of meaningless output. For those of us doing R&D in Taiwan, when facing these massive compute whales, us little guys really need to know how to protect ourselves.

---

2. Rare Joint Warning from NSA and Two Other US Agencies: Chinese AI Firms Launching "Industrial-Scale" Knowledge Distillation Plagiarism Against the US

• Source: CISA (https://www.cisa.gov/news-events/cybersecurity-advisories/aa26-251a）
• Summary: The US NSA, CISA, and FBI have issued a joint security advisory, directly naming six Chinese AI companies—DeepSeek, Moonshot AI, Alibaba, MiniMax, 01.AI, and StepFun. The advisory accuses them of extracting billions of tokens from US models like Claude and GPT through millions of requests since late 2024 for "knowledge distillation," emphasizing that this has become a core strategy for these Chinese AI firms.
• Most Surprising Part: US intelligence agencies directly accused them of "industrial-scale" plagiarism rather than sporadic data collection, pointing out that even DeepSeek's viral R1 and V3 models were trained on synthetic data generated by US models.
• Taiwan Perspective: As the global hub for semiconductors and AI server hardware, Taiwan must be highly vigilant about this new type of "software-level intellectual property" dispute, which could accelerate the complete decoupling of the US and China in the AI software ecosystem.
• Discussion Points:
  1. Does using APIs to perform "knowledge distillation" on a competitor's model count as legitimate technical learning?
  2. Will this report prompt the US to further restrict Chinese IP addresses from accessing Western AI APIs?
• Script Suggestion: Have you ever wondered how certain Chinese AI models managed to make massive leaps in performance overnight, while being so cheap they're practically free? US intelligence agencies have just laid their cards on the table, accusing six companies, including DeepSeek and Moonshot AI, of launching "industrial-scale" knowledge distillation. Simply put, they send massive armies of bots daily to spam GPT and Claude APIs, using other people's hard work as a free tutor to "distill" their own models. In Taiwan's tech industry, this looks like taking "copy-paste culture" to the absolute extreme. No wonder the US was mad enough to have the FBI and NSA team up to issue this warning. It looks like the AI Great Firewall will only get higher, and developers in Taiwan need to be extra careful about compliance risks when choosing model providers.

---

3. Meta Launches All-New Muse AI Assistant: It Can Check Your Mail, Book Trips, and Directly Swipe Your Card to Spend Your Money!

• Source: TechCrunch (https://techcrunch.com/2026/09/08/meta-debuts-its-muse-ai-agent-will-consumers-trust-it/）
• Summary: Meta has officially launched "Muse," a consumer-facing AI agent that integrates email reading, calendar management, and travel booking. It uses Stripe's Link mechanism to generate virtual card numbers and check out directly for users, offering three pricing tiers ranging from free to $100/month.
• Most Surprising Part: Meta actually has the nerve to ask users for access to their email inboxes and credit cards, even though engineers on the backend can still access this highly sensitive, underlying private data.
• Taiwan Perspective: Taiwanese consumers are highly sensitive to personal data privacy and online scams. Given that Meta has yet to fully resolve the rampant social media ad scams in Taiwan, getting Taiwanese users to hand over control of their credit cards and emails presents an incredibly high trust barrier.
• Discussion Points:
  1. Would you be willing to hand over your email password and credit card authorization to Meta just to save ten minutes booking flights and hotels?
  2. With AI agents directly handling financial transactions, will this trigger a new wave of "AI-driven credit card fraud"?
• Script Suggestion: Meta is making a massive leap this time—their new Muse assistant can actually swipe your card and buy things for you! It integrates Stripe's technology to automatically generate one-time virtual card numbers for checkout. Just imagine telling it, "Book me a flight to Tokyo next week," and it quietly handles the payment for you. But wait, this is Meta we're talking about! The same Meta with a history of endless privacy scandals now wants to read your inbox and hold your credit card. Even scarier, foreign media revealed that Meta's backend engineers can still see this private data. Personally, I would never hand my wallet over to it. With so many scams in Taiwan right now, if the AI gets "emotionally blackmailed" by a hacker, you might wake up to a completely maxed-out credit card!

---

4. European AI Giant Mistral Secures Record-Breaking €3 Billion Funding, Vows to Build Own Data Centers Instead of "Renting"

• Source: TechCrunch (https://techcrunch.com/2026/09/08/mistral-raises-e3b-as-sovereign-ai-becomes-big-business/）
• Summary: French AI startup Mistral has closed a massive €3 billion equity funding round, sending its valuation soaring to €21 billion. Led by Samsung Electronics, with participation from giants like NVIDIA, ASML, and BlackRock, Mistral plans to use the funds to build its own data centers instead of renting cloud services.
• Most Surprising Part: As Europe's flagship champion against OpenAI, Mistral's lead investor is actually South Korean tech giant Samsung, and they've decided to build their own data centers rather than renting cloud space.
• Taiwan Perspective: With Samsung leading and ASML participating, it's clear that "Sovereign AI" has become a hard-power geopolitical showdown. Taiwan's server supply chain and TSMC are poised to capture massive hardware orders in this wave of self-built data centers.
• Discussion Points:
  1. Why is Mistral insisting on "building its own data centers" instead of renting cloud capacity from Microsoft or Amazon?
  2. What kind of strategic layout does Samsung, as an Asian enterprise, signal by leading the investment in European Sovereign AI?
• Script Suggestion: Europe's top AI player, Mistral, made history today by raising a whopping €3 billion in one go! What's most interesting is that the lead investor is South Korea's Samsung. It seems Asian tech conglomerates want a piece of Europe's "Sovereign AI" pie too. And Mistral has real backbone—now that they have the cash, they don't plan on renting cloud servers from Microsoft or Amazon. They want to "build their own data centers"! It's like skipping renting and just buying land to build your own house. For our electronics contract manufacturers in Taiwan, this is an absolute jackpot! Whether it's server chassis, cooling chips, or TSMC's foundry services, this wave of "sovereign AI self-building" is going to bring massive profits to Taiwan's supply chain.

---

5. A Quiet Revolution Amidst the Noise! DeepMind Releases 1PB Genomic Atlas, Predicting the Consequences of 9 Billion Mutations Across the Human Genome

• Source: Nature (https://www.nature.com/articles/d41586-026-02835-4）
• Summary: Google DeepMind has released the AlphaGenome Atlas, a massive 1PB database that predicts the molecular consequences of all 9 billion possible single-DNA letter mutations in the human genome. This is over 30 times larger than the previous AlphaFold database and is being made freely available for non-commercial research.
• Most Surprising Part: While everyone is bickering over chatbots and academic gossip, this breakthrough—30 times larger than AlphaFold—is the "real AI technology" that can actually save the lives of countless rare disease patients.
• Taiwan Perspective: Taiwan possesses a highly robust National Health Insurance database and strong biomedical research capabilities. DeepMind open-sourcing this massive genomic atlas will create powerful synergies with Taiwan's precision medicine and targeted cancer drug development.
• Discussion Points:
  1. Why do AI breakthroughs with such massive contributions to human health often get less social media buzz than chatbots or gossip?
  2. What kind of disruptive innovation will open-sourcing such a massive genomic prediction database bring to the biotech and pharmaceutical industries?
• Script Suggestion: While everyone is arguing over AI gossip, Google DeepMind quietly dropped a real tech bombshell! They launched the "AlphaGenome Atlas," a massive 1PB database that predicts all "9 billion possible mutations" in the human genome. This is thirty times larger than AlphaFold, which shocked the world a few years ago! What does this mean? In the future, when doctors encounter rare diseases, they won't have to grope in the dark anymore—they can just look it up in this "genomic dictionary." Personally, compared to chatbots that just write code or chat with you, this kind of AI that can actually save lives is where technology is at its warmest and most beautiful.

---

6. Rather Self-Sabotage Than Face Fines? Google Admits EU Regulations Have Caused the Worst Degradation in Search Quality in 29 Years

• Source: Search Engine Land (https://searchengineland.com/google-says-dma-changes-in-eu-resulted-in-worse-degradation-of-search-quality-ever-487340）
• Summary: To comply with the EU's Digital Markets Act (DMA), Google modified how search results are displayed in Europe, removing convenient features like flight date filters and real-time pricing. Google publicly stated to users that this has led to the worst decline in search quality in the company's 29-year history.
• Most Surprising Part: Google actually publicly admitted, "Our product is now terrible to use," using this as a silent protest and public relations counterattack against EU regulators.
• Taiwan Perspective: As Taiwan explores antitrust and digital platform regulation bills, it must be wary of this side effect where "regulations lead to a regression in user experience." How to strike a balance between protecting market competition and maintaining product convenience is something Taiwanese lawmakers need to think deeply about.
• Discussion Points:
  1. Is Google's "self-sabotage" strategy truly about compliance, or is it trying to leverage user anger to pressure the EU?
  2. When regulations mandate "fair competition" but result in a worse consumer experience, is such oversight truly reasonable?
• Script Suggestion: This has got to be one of the most bizarre sights in tech history. Google is actually telling users across Europe, "We're sorry, but our search service is now terrible to use!" This isn't because engineers wrote bad code; it's to comply with the EU's Digital Markets Act. Google removed real-time flight comparison and hotel booking features directly from search pages, forcing traffic to third-party sites like Expedia. Google even bluntly stated that this is the worst degradation of search quality in their 29-year history. This move is incredibly clever. On the surface, they are playing by the rules, but in reality, they are throwing a tantrum at the EU, hoping angry users will protest. It also teaches us a lesson: sometimes, even when the government's intention is good "antitrust" protection, the ones who end up paying the price and suffering are still us ordinary folks.

---

Closing Thoughts

That's all for today's "Mark's Tech Insights"! From OpenAI's academic drama and Meta's bold experiment to DeepMind's life-saving genomic atlas, we are living in an era of rapid technological acceleration. I'm Muyan. See you next time, bye-bye!