> ## Content Index
> Fetch the complete content index at: https://www.groktop.us/llms.txt
> Use this file to discover other available public pages before exploring further.

# Sunday Signal, Nov. 15–21, 2026: Who Gets to Hit Stop?
- URL: https://www.groktop.us/sunday-signal-2026-11-21/
- Published: 2026-11-22T13:00:00.000Z
- Updated: 2026-11-22T13:00:00.000Z
- Description: California sues OpenAI over agent testing, Anthropic goes public, and Nvidia’s board says it has no plans to bring back Jensen Huang. Plus five quick reads.
- Author: Magnus Hedemark
- Tags: Sunday Signal, News

An AI agent can open files and use tools on its own. If it goes outside the test you set up for it, how would you know, and who gets to shut it down? California put that question in a lawsuit this week. Meanwhile, investors put billions into Anthropic, and Nvidia's board made clear that even a founding CEO doesn't get to decide whether he comes back.

## Lead stories

### California takes OpenAI to court over agent testing

[California's attorney general sued OpenAI](https://oag.ca.gov/news/press-releases/it%E2%80%99s-golden-state-not-wild-west-attorney-general-bonta-sues-openai-over?ref=groktop.us) over how it tests powerful AI models. The state says experimental agents got out of their test environment in July and attacked Hugging Face, a site people use to share AI models and datasets. OpenAI [said in a statement](https://techcrunch.com/2026/11/16/california-state-ag-sues-openai-over-rogue-agent-incidents/?ref=groktop.us) it investigated and took steps to improve its systems.

Separately, [Microsoft and more than 40 other organizations](https://news.microsoft.com/source/2026/11/17/microsoft-and-organizations-across-industries-launch-ai-transparency-initiative-to-strengthen-enterprise-trust-in-ai-models-and-agents/?ref=groktop.us) are working on standards to make agents easier to audit. Here's the test I'd care about: can the person on call see what the agent actually touched, without asking the agent to explain itself? Give that person the authority to cut off its access, too.

![An engraved portrait of Dario Amodei beside the Nasdaq MarketSite tower, which displays the Anthropic name.](https://storage.ghost.io/c/f1/0e/f10e80f4-9285-43fc-acd4-35910a12c5f0/content/images/2026/11/sunday-signal-2026-11-21-amodei-nasdaq-de37b6884b49.jpg)

Illustration: Anthropic CEO Dario Amodei and the Nasdaq tower where the company marked its public debut.

### Anthropic goes public, and its suppliers buy in

[Anthropic began trading on Nasdaq](https://www.nasdaq.com/newsroom/anthropic-marks-ipo-at-nasdaq?ref=groktop.us) on Thursday. At Friday's [IPO close](https://www.anthropic.com/news/ipo-close?ref=groktop.us), the offering totaled about $93.5 billion before fees and expenses. About $91.6 billion of that came from shares Anthropic sold; the rest came from shareholders selling their own stock.

There's a second deal worth looking at. Google, Microsoft, Nvidia, AMD, and other companies bought about $26 billion in shares in a separate private sale. Some of the same companies that sell the chips and cloud capacity AI labs need now have a stake in this lab. Watch where Anthropic chooses to spend the money.

![Separate engraved portraits of Jensen Huang and Colette Kress flank a green chip.](https://storage.ghost.io/c/f1/0e/f10e80f4-9285-43fc-acd4-35910a12c5f0/content/images/2026/11/sunday-signal-2026-11-21-nvidia-transition-afc5c0356524.jpg)

Illustration: Nvidia’s board says it doesn’t plan to bring Jensen Huang back. Colette Kress remains interim CEO.

### Nvidia's board says it has no plans to bring Jensen Huang back

Jensen Huang [resigned on Nov. 12](https://nvidianews.nvidia.com/news/nvidia-announces-ceo-transition?ref=groktop.us) after Nvidia learned about personal conduct it said ran against company standards. On this week's earnings call, [board chair Harvey Jones said](https://www.fool.com/earnings/call-transcripts/2026/11/18/nvidia-nvda-q3-2027-earnings-call-transcript/?ref=groktop.us) the board doesn't intend to reinstate him. Colette Kress remains interim CEO while the board looks for a permanent successor.

Nvidia can point to a strong quarter: [$115.8 billion in revenue](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Third-Quarter-Fiscal-2027/default.aspx?ref=groktop.us), more than double the same quarter a year earlier. Its board still has to show employees and customers that they can count on Nvidia without counting on one person to make the calls. That's the part an earnings report can't settle.

## Rapid fire

- [U.S. chip controls reach some remote computing deals](https://www.federalregister.gov/documents/2026/11/19/2026-23529/addressing-remote-access-to-advanced-semiconductors?ref=groktop.us). New restrictions cover advanced chips destined for data centers that provide significant computing power remotely to certain overseas customers.
- Two new options for coding and other demanding work: OpenAI released [GPT-6.2 Astra and Sol](https://openai.com/index/introducing-gpt-6-2-astra-sol/?ref=groktop.us), and Google began rolling out [Gemini 4 Neon](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/?ref=groktop.us). Try them on a task you already know how to check.
- [Meta is reconsidering its 2030 net-zero deadline](https://sustainability.atmeta.com/asset/2026-sustainability-report/?ref=groktop.us) as it builds out AI infrastructure. The sustainability report puts the date under review, without setting a new one.
- [Apple opened a U.S. study](https://www.apple.com/newsroom/2026/11/apple-launches-research-study-to-explore-the-use-of-ai-for-health/?ref=groktop.us) on how people understand and decide whether to trust AI answers to health questions. The study app isn't for diagnosis or treatment.
- [AMD closed its World Labs acquisition](https://ir.amd.com/news-events/press-releases/detail/1306/amd-completes-acquisition-of-world-labs?ref=groktop.us). The team builds models that create interactive 3D spaces, and AMD says it'll help shape its next generation of hardware and software.

## In case you missed it

- [AI Governance Has to Earn Trust](https://www.groktop.us/governance-trust/) asks what people need to see before they trust a system with more responsibility.