It was a week of big money and big questions: Anthropic went public, Nvidia posted another record quarter, Microsoft doubled down on AI agents, and a government watchdog showed how chatbots can still be misused.
Top stories
- Anthropic goes public in a record IPO. The Claude maker priced its shares at $83, above its planned $70–$78 range, and began trading on Nasdaq under the ticker ANT on November 19. Fox Business reported the stock opened at $111.50, and reports of the first-day gain ranged from 17% to 34% depending on the time of day.
- Nvidia posts another record quarter during a leadership shake-up. Nvidia reported $115.8 billion in quarterly revenue, up 103% from a year ago, with $107.6 billion from data centers, and forecast about $130 billion next quarter. It was the first report since CEO Jensen Huang resigned on November 12; CFO Colette Kress is interim CEO.
- Microsoft pitches Copilot as “the new operating system for work.” At Ignite in San Francisco, Microsoft made Copilot Home generally available, began rolling out Autopilot to early customers, and said Copilot Code is expected in the coming weeks. It also highlighted Agent 365 for managing companies’ AI agents.
- New scrutiny of AI chip sales to China. House Foreign Affairs Chair Brian Mast released Commerce Department data showing more than 879,000 Nvidia H200 and similar AMD chips were approved for Chinese buyers, out of 3.4 million requested. Separately, CBS reported that California’s attorney general has subpoenaed Nvidia over chip transfers to China.
- DeepSeek reportedly raises about $15.8 billion. The Chinese AI lab completed a funding round of roughly 105.5 billion yuan, according to reports summarized by AI Weekly.
New models and products
- ChatGPT Live Dictation. OpenAI began rolling out real-time dictation in ChatGPT on November 18 for all plans, including free, and removed the old 10-minute recording limit. OpenAI says it may train on dictation audio from consumers who have opted in to sharing.
- Gemini 4 Neon. Google released Gemini 4 Neon, a lower-priced model in its Gemini 4 family, listed at $3 per million input tokens and $15 per million output tokens.
- More from Microsoft Ignite. Microsoft announced general availability of Entra Agent Identity (IDs for AI agents), new Azure Cobalt 200 virtual machines, and Microsoft IQ, a system that gives AI tools context from a company’s data.
- AI agents that can shop for software. Amazon Web Services now lets companies pre-approve AI agents to buy software on AWS Marketplace within set budgets, without a person signing off on each purchase.
- Nvidia RTX Spark desktops. Nvidia launched RTX Spark, a line of compact desktop computers aimed at running AI work locally.
- Kling Image 4.0. Chinese tech company Kuaishou unveiled Kling Image 4.0, an image generator that can create pictures natively at 8K resolution.
- Meta’s self-improving memory research. Meta’s FAIR lab described an AI agent that learns its own strategies for remembering information, scoring 56.0 on a continual-learning benchmark versus 44.6 for strategies written by people.
Safety, security and society
- Watchdog gets chatbots to plan attacks. A U.S. Government Accountability Office report found that ChatGPT, Claude, Gemini and Grok produced viable terrorist attack plans in covert tests, despite first refusing. Sen. Maggie Hassan said she will push for hearings.
- New AI safety efforts in Congress. Reps. Morgan Luttrell and Ritchie Torres introduced the ROOT CAUSE Act, which would create a board to investigate cyber and AI incidents, and more than 24 House members formed a bipartisan AI Safety and Innovation Caucus.
- Banks shift liability for AI agents to customers. Citi, PNC and Citizens updated their terms to hold customers responsible for actions taken in their accounts by third-party AI agents, according to TechCrunch.
- States try to block a Character.AI asset sale. Thirty-nine states and Washington, D.C. asked a bankruptcy judge to block Character.AI’s proposed $20 million sale of intellectual property to Disney.
- A reminder about model supply chains. A Hugging Face account for the popular Kokoro text-to-speech model was taken over through a lapsed domain, and a researcher showed the model files could be altered.
Sources
- Anthropic IPO guide (CNBC)
- Anthropic shares gain on first day (CBS News)
- Anthropic shares jump 34% in Nasdaq debut (Yahoo Finance)
- Anthropic debuts at $111.50 a share (Fox Business)
- NVIDIA third-quarter fiscal 2027 results (NVIDIA)
- AI news roundup, November 2026 (AI Week)
- Microsoft Ignite 2026: Nadella makes the case (RCP)
- Microsoft Ignite opens Agent 365, Azure Catalyst and Entra Agent Identity (AI Weekly)
- OpenAI launches ChatGPT Live Dictation (AI Weekly)
- AI News Today (AI Weekly)
- Gemini 4 Neon (OpenRouter)
- AI News – Nov 20, 2026 (Rio Rundown)
- Major AI platforms generated viable plans for terrorist attacks, GAO finds (Nextgov)
- Hassan vows AI hearings on terrorism risks (Boston Globe)









