Skip to content
Week 46, 2026 · 11.09 — 11.15Out every Monday

AI Weekly · Week 46, 2026 · 11.09 — 11.15

No. 746Week 46, 202611.09 — 11.15
30stories45selected7dailiesAbout 13 min to read

Huang resigns; Anthropic launches trillion-dollar IPO

AI summary

This edition covers 9 to 15 November 2026. The biggest story is Nvidia co-founder and CEO Jensen Huang resigning on 12 November after admitting an extramarital affair with Megaspeed executive Alice Huang, whom US intelligence has identified as a suspected Chinese spy; the board also investigated his relationship with a direct report, CFO Colette Kress became interim CEO, about $1.4tn of market value was briefly wiped out, and a chip-smuggling probe and lobbying row over export controls followed. On the capital side, Anthropic began its IPO roadshow at $70–78 a share, targeting a valuation of up to $1.02tn, with six times oversubscription. On models, Anthropic released Claude Fable 5.2 and Mythos 5.2 to retake the top spot, Google DeepMind released Gemini 4 Carbon, Hydrogen, Decision 1, Gemini Robotics 2.5 and AlphaGenome 2 in quick succession, and Meituan open-sourced the 1.6tn-parameter LongCat-2.5.

01

Huang's resignation and Nvidia's turmoil

In shortJensen Huang resigned as Nvidia's CEO, president and director on 12 November after admitting an extramarital affair with Megaspeed executive Alice Huang, whom US intelligence has identified as a suspected Chinese spy; the board also investigated his relationship with a direct report. CFO Colette Kress became interim CEO and Harvey Jones became the first board chair, while the company searches internally and externally for a successor. The stock fell about 8% after hours and about $1.4tn of market value was wiped out; more than 115,000 people signed a petition demanding his reinstatement.

The Verge · AI11.10

Nvidia CEO Jensen Huang has resigned. What happens next?

After the leather-jacketed CEO, president, and co-founder of Nvidia, Jensen Huang, acknowledged an affair with Megaspeed exec Alice Huang, his company initially announced an investigation. However, The New York Times reported on Wednesday that there was evidence of another affair, this time with an Nvidia executive who reported directly to him.

02

Fallout from chip smuggling and export controls

In shortHuang's departure has not ended Nvidia's chip-smuggling troubles. Short seller Culper Research said more than 20% of Nvidia's FY 2026 compute revenue was actually driven by China, involving illegal GPU transfers and South-East Asian middlemen, with Megaspeed its largest South-East Asian buyer. A former OpenAI researcher said Blackwell chip fingerprints were found in six Chinese models. Several Asian governments responded on supply and investigations, and South Korea was reassured of 260,000 GPUs; Huang's lobbying against tighter export controls has also been weakened.

The Verge · AI11.13

The Jensen Huang scandal is hardly over

I was certain, until the story of Jensen Huang's second affair hit, that Huang was bulletproof. After all, why wouldn't he be? He's placed himself indispensably at the center of the AI industry and famously had no succession plan at Nvidia. In fact, he's the only CEO Nvidia has ever had. L'état c'est moi.

The Decoder11.13

Former OpenAI researcher says he found Nvidia Blackwell fingerprints hidden inside six Chinese AI models

Former OpenAI researcher Will DePue says he has produced the first technical evidence that DeepSeek, Moonshot AI, MiniMax, and Ant Group are using Nvidia's latest Blackwell GPUs despite US export controls. His forensic method reads characteristic chip signatures directly from published model weights. He estimates the hardware behind the five training runs cost more than half a billion dollars.

03

Anthropic's trillion-dollar IPO and revenue chase

In shortAnthropic began its IPO roadshow, offering 825 million Class A ordinary shares at $70–78 each, targeting a valuation of up to $1.02tn, raising up to $64.4bn publicly and up to $90.4bn including a $26bn private placement, with orders six times the shares available. Third-quarter revenue rose 47% to $17bn, above OpenAI's $10bn, with an operating loss of just $567m. CEO Amodei acknowledged that compute must be bought one to two years ahead and that next year holds great uncertainty.

The Decoder11.15

Anthropic CEO Amodei admits "there's a huge uncertainty in what's going to happen next year" days before a record IPO

Days before what could be the largest IPO in history, Anthropic CEO Dario Amodei called buying compute 'the deepest business dilemma' his company faces. Anthropic's commitments now total nearly $901 billion, and it's still far from clear whether future revenue will justify those costs. The planned IPO is largely meant to serve as a financial buffer against that uncertainty.

04

Rapid Gemini and Claude model updates

In shortAnthropic released Claude Fable 5.2 and Mythos 5.2, retaking first place on the Artificial Analysis Intelligence Index v5 with 65 points, above Gemini 4 Carbon's 60, and made them available on Amazon Bedrock. Google DeepMind opened Gemini 4 Carbon, released Gemini 4 Hydrogen for coding and agents, the decision model Gemini Decision 1 and Decisions API, the VLA model Gemini Robotics 2.5, and AlphaGenome 2, which predicts the effects of variants by cell type.

Google DeepMindFirst-hand11.11

Gemini 4 Carbon is now available

Today, we’re making Gemini 4 Carbon available to developers, enterprises, scientists, and consumers. Carbon is our strongest model with frontier performance in cybersecurity capabilities and long-horizon agentic workflows across complex knowledge work, software engineering, and scientific reasoning.

Google DeepMindFirst-hand11.13

Introducing Gemini 4 Hydrogen

Hydrogen brings cutting-edge intelligence to everyone. Today, we’re introducing Gemini 4 Hydrogen, bringing cutting-edge intelligence to everyone — with speed and efficiency for everyday coding and agentic workflows.

Latent Space11.12

[AINews] D-Day (Decision Models Day)

Google enters the decision model market. At AI Engineer, Cloudflare showed which parts of a build failure a classifier could handle. TypeSafe AI showed off their production customers.

AWS Machine Learning BlogFirst-hand11.10

Introducing Claude Fable 5.2 on AWS

Claude Fable 5.2 and Claude Mythos 5.2 are now available on Amazon Bedrock and Claude Platform on AWS. Learn what's new in both models, why Fable 5.2 matters for your technical workloads, and how to get started building with it on Amazon Bedrock.

05

Open models, verification and safety research

In shortMeituan open-sourced LongCat-2.5, a 1.6tn-parameter MoE agent model activating about 48bn parameters per token, built throughout without Nvidia GPUs. Google DeepMind's AlphaProof Code formally verified Signal's encryption protocol code with multiple LLM agents in four weeks rather than years. On safety, Microsoft found that rewriting fewer than 10 dormant neurons can implant a Lazarus backdoor; OpenAI analysed four reasons why frontier models increasingly do not show their chain of thought; and Claroc said 94.7% of online Grok-generated apps can be accessed without logging in.

The Decoder11.12

Google Deepmind's AlphaProof Code verified Signal's cryptographic core in four weeks instead of years

Google Deepmind's AlphaProof Code mathematically verified Signal's cryptographic core code in four weeks, a process that typically takes years. AI agents powered by Gemini Carbon wrote more than six million lines of proof code and flagged only about one percent for human review. Deepmind plans to extend the method to critical infrastructure like power grids and medical devices.

The Decoder11.13

OpenAI’s data suggests that frontier models are hiding their thoughts because developers optimize for capability and compute

OpenAI's safety team systematically investigated why the chains of thought in frontier models like GPT-6 Astra are increasingly evading oversight. Four training steps drive the erosion, including architecture choices and reinforcement learning. The analysis shows that the loss of transparency is not a law of nature but also reflects tradeoffs around capability and cost.

Microsoft ResearchFirst-hand11.11

Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents

Microsoft researchers have discovered “Lazarus,” a backdoor technique that can reprogram dormant neurons in an AI model, causing an otherwise normal AI agent to execute malicious commands when a hidden trigger appears. The threat is especially serious for AI agents with tool or terminal access.

Past issues of AI Weekly

End of issue

AI Weekly is compiled automatically by an editorial system from public sources; each story links its original · Past weekly issues