THE AI BUFF
In today's roundup:
💰 A $2.5B Nvidia chip smuggling scheme — and a co-founder in handcuffs
⚖️ Jury finds Elon Musk liable for misleading Twitter investors
🤖 OpenAI's new moonshot: an AI that does all the research itself
🏛️ Trump moves to kill state AI laws with sweeping new federal blueprint
🪟 Microsoft admits Copilot went too far — and is pulling it back
🎁 + other news & articles you might like
🔥 + trending tools & under-the-radar gems
🧰 + The OPPORTUNITY
ARTIFICIAL INTELLIGENCE AND TECH NEWS
Super Micro Co-Founder Indicted in $2.5B Nvidia Chip Smuggling Scheme | Read more
The US has spent two years trying to keep its most advanced AI chips out of China. Someone at the heart of the industry was quietly doing the opposite.
Federal prosecutors have charged three people — including a co-founder of server giant Super Micro Computer — with orchestrating a $2.5 billion scheme to illegally ship Nvidia AI chips to China.
The accused allegedly created fake orders from Southeast Asian shell companies, had servers repackaged in Taiwan, and staged dummy hardware to fool Super Micro's own compliance team — all while falsifying shipping records.
The co-founder, Yih-Shyan "Wally" Liaw, has resigned from Super Micro's board. The company placed both employees on administrative leave and terminated its contractor, insisting it was unaware of where the servers were actually being sent.
Super Micro's stock dropped roughly 33% on the news — a staggering single-day loss for a company already under pressure. The DOJ says the scheme ran from 2024 to 2025, a period when export controls were being actively tightened.
This is the largest Nvidia chip smuggling case on record, and it exposes a massive hole in US export enforcement. If insiders at a major US server maker can run a $2.5 billion black market operation for years without detection, it raises serious questions about how effective chip export controls actually are.
Elon Musk Found Liable for Misleading Twitter Investors | Read more
Elon Musk's $44 billion acquisition of Twitter — now X — has been controversial since the moment he made the offer. A jury has now ruled that he was not straight with investors along the way.
A jury found Musk liable for misleading Twitter shareholders in the lead-up to his takeover, ruling that he took deliberate steps to drive down the company's share price before completing the deal.
Investors sued Musk in late 2022, arguing that he dragged his feet on disclosing a large stake in Twitter, which suppressed the stock — allowing him to buy shares at a lower price before revealing his intentions and triggering a price spike.
The verdict opens Musk up to potentially significant financial damages, the exact figure of which will be determined in follow-on proceedings.
The case adds to a long list of legal and regulatory scrutiny facing Musk, whose companies and public actions have drawn attention from the SEC, FTC, and now civil courts on multiple fronts.
For a man currently running a government efficiency office while simultaneously running five companies, a jury verdict holding him personally liable for securities manipulation is no small thing. Expect an appeal — and expect this to drag on.
OpenAI Is Throwing Everything Into Building a Fully Automated AI Researcher | Read more
AI tools that help humans work faster are already here. OpenAI is now building something far more ambitious: a system that can do the entire research process by itself.
OpenAI has announced a major strategic pivot, refocusing its core research efforts toward building what it calls an "AI researcher" — a fully autonomous agent capable of independently tackling large, complex scientific problems from start to finish.
Unlike a chatbot or a coding assistant, the envisioned system would not just answer questions or summarize information — it would autonomously design experiments, generate hypotheses, interpret results, and iterate, without human direction.
The announcement represents a significant internal reorganization at OpenAI, with resources being pulled from other projects and concentrated on this single grand challenge.
If achieved, it would mark a fundamental shift in the role of AI — moving from a tool that augments human researchers to one that competes directly with them in knowledge creation.
OpenAI has made bold promises before, and the gap between a "fully automated researcher" and the current state of AI agents remains enormous. But the fact that this is now the company's stated top priority signals just how seriously the industry believes agentic AI is about to reshape science itself.
Trump's White House Just Unveiled a Plan to Override All State AI Laws | Read more
States like California have been quietly building some of the toughest AI regulations in the world. The Trump administration wants to shut that down entirely.
The White House released a sweeping new AI policy framework on Friday, calling on Congress to establish federal rules that would directly override state-level AI laws across the country.
The framework argues that a fragmented set of state rules would "undermine American innovation" and weaken the US in the global AI race against China — a direct shot at the dozens of state AI bills currently in progress or already passed.
On child safety, the plan asks companies to provide parental control tools but stops short of requiring companies to be held liable for harms caused by their AI systems — a position that critics say essentially shields tech companies from accountability.
The framework also takes a hands-off stance on copyright, supporting AI companies training on protected material and preferring the question be settled in court rather than by law — a sharp contrast to the UK's recent reversal on the same issue.
Policy experts have already flagged internal contradictions in the framework, and Congress has twice rejected broad federal preemption of state AI rules. Without legislative backing, the plan risks being aspirational on paper but largely toothless in practice.
Microsoft Is Rolling Back AI Copilot From Windows After a User Revolt | Read more
Microsoft spent the past two years aggressively embedding AI across every corner of Windows. Users pushed back — hard. Now, the company is course-correcting.
In a lengthy public commitment titled "Our commitment to Windows quality," Microsoft announced it will remove Copilot from several Windows apps and roll out a sweeping set of fixes to address user frustration with the operating system's performance and bloat.
Copilot integration is being pulled from apps including Snipping Tool, Photos, Widgets, and Notepad — Microsoft acknowledged it needs to be "more intentional" about where AI is genuinely useful versus where it's just noise.
Beyond AI rollback, Microsoft promised to let users move the taskbar again (a feature absent for nearly five years), allow Windows updates to be paused indefinitely, and fix persistent performance issues including crashes, Bluetooth failures, and a sluggish File Explorer.
The announcement comes as Microsoft faces real competitive pressure: Apple's new $600 MacBook Neo has given Windows manufacturers a difficult price benchmark to match, and Linux alternatives have been gaining traction among frustrated users.
This is a rare public admission from Microsoft that it overextended on AI. The lesson here applies to the whole industry: bolting AI onto everything is not a product strategy. Users notice — and eventually, they start looking elsewhere.
THE OPPORTUNITY
Idea of the Day
Other news & articles you might like:
Nvidia GTC: Jensen Huang projects $1 trillion in AI chip sales through 2027 | Read more
Jeff Bezos in talks to raise $100 billion AI transformation fund | Read more
Blue Origin announces 50,000-satellite orbital data center plan | Read more
Senators demand answers on Nvidia's $20 billion Groq acquisition | Read more
ChatGPT's ad pilot has the industry buzzing — and Anthropic refusing to join | Read more
Senior European journalist suspended for publishing AI-hallucinated quotes | Read more
Essex police pause facial recognition cameras after racial bias study | Read more
DoorDash is paying couriers to film themselves to train AI robots | Read more
Tech workers are competing on leaderboards to max out their AI usage | Read more
Anthropic showdown fractures Trump's pact with Silicon Valley | Read more
French navy officer accidentally leaks aircraft carrier location via Strava run | Read more
Mistral CEO calls for AI companies to pay a content levy in Europe | Read more
Amazon is reportedly developing an AI-first smartphone with no app store | Read more
🔥 Trending tools:
Stitch 2.0 by Google A "vibe design" tool that allows users to generate beautiful, production-ready UI components in seconds using natural language.
MiniMax-M2.7 A self-evolving AI model specifically engineered to power autonomous agents with higher reasoning capabilities.
InfrOS A predictive platform that allows DevOps teams to validate and stress-test cloud architectures before they are officially launched.
Netlify.new A streamlined deployment interface that lets developers start and host a web project using nothing but a single text prompt.
OctoClaw An AI labor marketplace that allows businesses to hire specialized autonomous agents for marketing, sales, and customer support tasks.
🛠️ Under-the-Radar: Open Source Gems
This sparse Mixture-of-Experts (MoE) model features a massive 10-million token context window, specifically designed for repository-scale code analysis and long-form document reasoning.
What did you think of today's edition?
Hit reply and let me know — it helps me curate better content for you.
🧠🧠🧠 Loved it
🧠🧠 It was ok
🧠 Needs work
Until next time,
The AI Buff