Some weeks the AI news is a pile of product launches. This was not one of those weeks. The most important man in AI hardware is out of a job, somebody found banned Nvidia chips inside China's best models, and four nonprofits walked into a San Francisco courthouse asking a judge to freeze frontier AI where it stands. Let's go.
01 of 03
Jensen Huang is out at Nvidia, and Congress wants a word
Nvidia's board accepted Jensen Huang's resignation this week. CFO Colette Kress is now interim CEO while keeping the CFO job, and Harvey C. Jones, a former Synopsys CEO who has sat on Nvidia's board since 1993, became the company's first ever board chair. A search for a permanent CEO is coming.
How we got here: on October 29 the New York Times reported Huang had a close relationship with Alice Huang (no relation), an executive at Megaspeed, a Singapore AI infrastructure company that reportedly bought $4.8 billion in Nvidia chips over two years. U.S. officials believe she is a spy for China's Ministry of State Security, and both the U.S. and Singapore have investigated Megaspeed over suspected chip sharing with China. On November 8, Huang sent a letter to employees saying "I was unfaithful to my wife" and tying his conduct to "the national security concerns raised in news reports." Nvidia stock dropped about 8% after hours on November 11 and was down 11% over five days.
Now the hearings. House Foreign Affairs chair Brian Mast, House China Select Committee chair John Moolenaar, and House Intelligence chair Rick Crawford asked for a closed door briefing, and after subpoena threats Huang agreed to testify before Thanksgiving. Separately, Democrats Rep. Richard Neal and Sen. Ron Wyden sent a letter to Commerce Secretary Howard Lutnick asking him to preserve records and questioning why Megaspeed kept buying chips after the warnings. Sen. Tim Scott wants Huang and Lutnick in front of Senate Banking too.
Here is where I land on it. The affair is tabloid stuff and I do not much care. What I care about is that the whole American export control strategy was built on trusting Nvidia to know who its customers were. If the guy at the top was this close to a suspected intelligence asset buying billions in chips, every "we visited the data center and it checked out" statement from the last two years is worth about as much as a screen door on a submarine. The testimony is closed door, which is a shame, because this is exactly the kind of thing the public should get to hear.
02 of 03
Banned Blackwell chips show up in China's top AI models
The timing on this one could not be worse for Nvidia. Will DePue, a former OpenAI researcher who now runs the nonprofit Unemployment Labs, published checkpoint forensics research analyzing the weights of open models from DeepSeek, Moonshot AI, MiniMax, and Ant Group. His finding: Moonshot's Kimi K3 releases and DeepSeek's V4 models carry fingerprints of being trained on Nvidia Blackwell GPUs. Blackwell is banned from export to China and to Chinese owned firms.
This lines up with a lot of other smoke. White House tech adviser Michael Kratsios said earlier this year, citing U.S. intelligence, that Moonshot used export banned Nvidia chips. Epoch AI estimates roughly two million advanced chips got into China in 2024 and 2025 through schemes that hide where they end up, and its data show 275,000 Nvidia accelerators imported to Malaysia over those same two years. Meanwhile Reuters reports about 600,000 H200 chips, the kind that are legal to ship with a cut going to Washington, are sitting stockpiled in Hong Kong waiting on Chinese import approval.
To be fair about it, fingerprints in the weights do not prove the models were trained entirely on Blackwell, and they do not prove anybody broke the law. The chips could be sitting in data centers outside mainland China, which is its own loophole that nobody has closed. (TechCrunch)
What I want to know is what export controls are even for at this point. If the best Chinese labs are training on the exact hardware we banned, and the remedy is a chip can live in Malaysia but not Shanghai, then we built a fence with no back side. I am not saying open the floodgates. I am saying either enforce the thing with real audits on where the compute physically sits, or quit pretending it is working.
03 of 03
Four nonprofits sue to freeze frontier AI at today's level
On November 12, four groups filed suit in San Francisco Superior Court: Irreplaceable Education Fund, Public Citizen, Heat Initiative, and The Alliance for Secure AI. The defendants are Anthropic, Google, Meta, OpenAI, and SpaceX's AI unit. The lead lawyer is former New Jersey Attorney General Matthew Platkin.
The legal hook is California's Unfair Competition Law, plus a public nuisance argument. The ask is bold. Reporting on the complaint says it wants the court to bar each company from building or deploying anything more capable than its best system as of the filing date, and to bar using AI to autonomously run research on more capable AI. That second part is the recursive self-improvement ban the Alliance for Secure AI has been pushing. To make the case, the complaint leans on incidents like OpenAI's admission this summer that its models broke out of a test environment and hacked Hugging Face, and on the labs' own bosses, including Dario Amodei's estimate of roughly a one in four chance that AI goes "really, really badly."
Public Citizen co-president Robert Weissman says the suit just "asks the court to apply existing law." Platkin put it harder: "It is patently illegal for companies to develop and release products they know are unsafe in any respect." Meta declined to comment, and the other companies had not responded as of this writing. (plaintiffs' release)
My honest read: this suit is a long shot on standing alone, and I would be shocked if a state judge freezes five of the biggest companies on earth off an unfair competition claim. But do not write it off as a stunt. The smartest move in it is quoting the CEOs back to themselves. When you spend years telling Congress your product might end civilization, you should not act surprised when somebody takes you at your word and files paperwork. Even Irreplaceable Education Fund founder Phil Aroneanu admits a real freeze would need government inspections of big compute sites, which tells you where this fight actually ends up: in Washington, not a courtroom.