Skip to content

Daily AI digest, November 18

Hands type on a glowing red typewriter as pages with a GPU, bell, chart, server rack and robot hand fly out News

The biggest AI news of November 18 came from Anthropic and Nvidia. Anthropic priced its IPO at $83 a share, above its range, and starts trading on Thursday. Nvidia stopped shipping H200 chips to China, and its board said it will not bring back Jensen Huang. Google released Gemini 4 Neon and opened Genie 4, and California set out new AI safety rules. Below are the big stories of Wednesday, November 18, plus the day's rumours.

Anthropic prices its IPO at $83

Anthropic priced its IPO at $83 a share, above the $70 to $78 range, a person familiar with the deal told CNBC. The offering was upsized to 979.6 million shares and raises about $81.3B. Anthropic had not announced the price itself by Wednesday evening.

The price puts Anthropic's value above the $1.4 trillion it had targeted, CNBC reports. Trading on Nasdaq under the ticker ANT is set to start on Thursday, November 19.

The same day, the Center for AI Safety and a coalition of researchers published a statement. It says Anthropic is not a safe AI company and accuses Anthropic of weakening its Responsible Scaling Policy and lobbying against state AI laws. The page carries no response from Anthropic.

Nvidia stops H200 shipments to China

Interim CEO Colette Kress said on the earnings call that Nvidia is no longer shipping licensed Hopper data center chips to China. Nvidia sold $2.4B of them in the last quarter, and its forecast counts on none. Our breakdown is here.

Board chair Harvey Jones said on the same call that the board "does not intend to reinstate Jensen as CEO", Al Jazeera reports. Nvidia shares rose 4.8% after hours on Tuesday.

At an all-hands meeting on Wednesday, Kress told staff to move forward and mentioned pay raises, CNBC reports, citing attendees. A ransomware gang called The Gentlemen also listed Nvidia on its leak site, though Nvidia has not confirmed a breach. More here.

House Democrats asked Cantor Fitzgerald for records on whether Commerce Secretary Howard Lutnick's export decisions were tied to its Nvidia holdings. Cantor called the letter inaccurate, CNBC reports.

Google releases Gemini 4 Neon and opens Genie 4

Google released Gemini 4 Neon, a fast reasoning model close to Gemini 4 Carbon in quality. It costs $3 per million input tokens and $15 per million output tokens. It is the fourth Gemini 4 model in five weeks. Details are here.

Google DeepMind also opened Genie 4, its newest world model, to Google AI Ultra subscribers in Project Genie. Users can walk through generated worlds, change them with prompts and, for the first time, hear them. More in our article.

California sets out next AI safety rules

Governor Gavin Newsom's office released recommendations for mandatory AI safety standards. They include independent testing, wider incident reporting and state power to order an emergency shutdown of a dangerous model. The plan would also let the state cap recursive self-improvement.

These are recommendations for now, building on SB 53, SB 813 and AB 1405. California also asked Congress and President Trump to use the framework as a floor for national rules.

A Microsoft package on PyPI carried a credential stealer

Two malicious versions of Microsoft's agent-framework Python package were published to PyPI on Tuesday evening, Wiz reports. They carried a credential stealer aimed at Azure, CI and cloud secrets. Both were removed within about an hour, and version 1.26.0 is the last safe one.

Wiz says the attacker appears to have compromised a Microsoft code-signing certificate. Microsoft has not commented on the page.

Smaller updates

Fifteen smaller stories came out on Tuesday and Wednesday.

  • Palo Alto Networks agreed to buy Applied Compute, an AI post-training startup, for $4.5B in stock, SiliconANGLE reports.
  • Isomorphic Labs, the Google DeepMind drug discovery spinout, raised a $5B Series C co-led by Greenoaks and Sequoia, TechCrunch reports.
  • Microsoft rebuilt Copilot at Ignite around Autopilot, an always-on agent with its own identity, now in early access.
  • Microsoft also released MAI Realtime 1, its first speech-to-speech model, at $0.01 per audio minute. Details are here.
  • GitHub added Adaptive reviews, cross-repository context and secret detection to Copilot code review. More here.
  • Replit made its Microsoft Fabric, Windows and Teams integrations generally available at Ignite. We covered it here.
  • OpenAI launched ChatGPT for Sales, with a Sales plugin and more than 30 CRM partner plugins. Details are here.
  • Anthropic added an experimental Code Mode to Claude Code, where Claude chains tools in a short JavaScript program. More here.
  • Perplexity launched the Orchestrator API, one endpoint for 45 models at list price with no platform fee. We covered it here.
  • Google AI Studio now sells custom domains for apps built with Gemini. Details are here.
  • Arena launched the Agent Arena Index, where Claude Fable 5.2 leads with 1,463 points. More in our article.
  • Tencent released the open weights of Hy4 flash on Hugging Face, two days after its API launch. Details are here.
  • Hub released HubEgo-100K, an open dataset of first-person video of skilled manual work for robot training. More here.
  • A China Telecom-led team released Xing4, a model family trained on 8,192 Huawei Ascend chips without any GPUs.
  • We also published a guide to the best AI video models for motion graphics, which is here.

Eight of the fifteen are tools for developers and companies building with agents.

Rumours

Three more stories circulated on November 18, but we could not read the primary report or confirm them with the companies.

  • Sierra, Bret Taylor's agent startup, is reportedly in talks to raise about $1.5B at a $30B valuation, The Information reports.
  • AI found five remote code execution flaws in Nginx for about $1,000 in tokens, and F5 patched two of them, Forbes reports. We could not read the report.
  • Google has reportedly made Gemini's Tasks feature free in six countries, The Verge reports.

Treat all three as unconfirmed until the companies announce them.

What the day was about

Wednesday put a price on Anthropic. Investors paid above the range for one of the largest listings on record, even as safety researchers challenged the company in public.

Nvidia, meanwhile, closed doors on two fronts: China sales and Huang's return. Google kept shipping models, and California pushed for rules that could shut down dangerous ones. On November 18 the money for AI reached a new high, and the demands on the companies grew with it.

Sources

Portrait of Felix Laurent, author at thehype.news

Felix Laurent

News Writer at thehype. Felix is a journalist with experience in marketing. He covers artificial intelligence and emerging technologies, focusing on the ideas, innovations, and trends shaping the future of tech.

More articles by Felix Laurent →
On air · radio.thehype.news ↗ AI news radio — 24/7