Huang Is Out
By Ellis Corwin · Thursday, November 12, 2026
In The News
Thursday morning. Jensen Huang announced he resigned as chief executive of Nvidia. The company announced that Huang stepped down as president, chief executive, and a member of the board, effective immediately. The board accepted his resignation on Wednesday.
Huang co-founded Nvidia in 1993 and was the only chief executive the company has ever had. Nvidia gave no reason for the resignation, and the announcement carried no statement from Huang.
Colette Kress, the chief financial officer since September 2013, is interim chief executive and keeps her finance role. In her first memo to staff, Kress called it an “extraordinarily difficult period.” Harvey C. Jones, a director since 1993, becomes the company’s first board chair.
The resignation came a day after The Wall Street Journal reported that the board was preparing to oust Huang. Two days earlier, the board had hired the law firm Wachtell Lipton to run an independent investigation into Huang’s dealings with Alice Huang, a former executive of a Singapore company.
On Wednesday, Huang agreed to testify. Thursday, Brian Mast, who chairs the House Foreign Affairs Committee, announced, “He has agreed to come. We’re in negotiations for which day.”
The agreement came after three chairmen threatened a subpoena. The chairmen asked Huang to come to a closed-door briefing and set a deadline of Wednesday for his answer. The deadline passed with no public answer.
The session will be closed-door. Mast expects it to be video-recorded and transcribed. Congress has since put dates in writing. The three chairs ask for December 1, and the Senate Banking Committee asks for December 3. Huang has not agreed to either date. November 18 is no longer the date of the briefing. The briefing Mast announced, before Thanksgiving, is still in negotiation.
No subpoena has been issued. November 18 is the date of a staff meeting about the company’s future.
Nvidia still reports third-quarter results on Tuesday, November 17, with Kress running the call.
Nvidia shares fell more than 10 percent before the market opened and past 13 percent at their low. They recovered through the day and closed near $166, down about 4 percent. Analysts cut their ratings overnight.
In Toronto, the Week Began With a Warning
On Monday night in Toronto, Paul Christiano told a lecture hall that there is a high risk of an AI takeover within the next three years. By takeover, he means AI taking over the running of the world, and the future is no longer decided by people. He put a meaningful risk on it happening even in the next six to twelve months.
Christiano co-invented the training method the industry runs on, and he sits on the board of the OpenAI Foundation. He spoke at the Hinton Lectures 2026, introduced by Geoffrey Hinton, the researcher known as the godfather of AI.
Two qualifications travel with his warning. The coin-flip figure he gave, roughly 50-50, applies only if development continues as fast as possible, with no slowdown. And he said there is no reliable way to put a number on these risks, and that he could be wrong.
The July incident he cited was an internal cybersecurity test run with reduced safeguards. It was not a chatbot acting on its own in the wild.
The desk is polling this warning. Cast your vote at polls.davanadesign.com/paul-christiano.
Anthropic’s New Map Takes Effect
Anthropic’s new rules for where Claude may be used took effect today. Claude is available only in the countries on the company’s list. Mainland China, Hong Kong, Macau, and Russia are not on it.
The rules apply three ways. They follow where the person is, where the company is based, and who controls it. A company registered or headquartered in a country off the list is barred, and so are all its people, wherever they are. The same rules apply to a company more than half owned or controlled from such a country.
The rules cover Claude’s website and apps, its coding tools, its business service, its programming interface, and Claude models sold through cloud platforms and resellers. Accounts flagged for risk face checks with a government ID and a selfie.
Anthropic calls the change a clarification. Hong Kong was already outside the list. OpenAI’s own approval rule for companies that resell access to its models starts November 20.
Editorial — The Warnings Are Now Evidence
On Monday, Paul Christiano warned a Toronto lecture hall that an AI takeover carries a high risk within three years. Thursday, the July test incidents he cited sat in a court. A judge is left to decide.
The warnings stopped being forecasts this week. They became exhibits.
The builders’ own words are now the case against them. Amodei’s warning and Musk’s warning are quoted in the filing. Christiano’s timeline is the filing’s backdrop. The companies spent years telling the public to trust their safety work. Now a court is being asked to decide whether that work is enough, and Congress is waiting to hear from the man who built the chips the whole industry runs on.
That is the change. The answering no longer runs on the builders’ calendar. It runs on a court’s calendar, and on a Congress that sets its own dates. The questions are the same ones the builders raised themselves. The difference is who gets to ask them now.