- Google DeepMind has made Gemini 4 Carbon available to developers, enterprises, scientists and consumers, which it calls its strongest model.
- Anthropic has launched the roadshow for an IPO of 825 million Class A shares priced at $70 to $78 each.
- OpenAI's revenue has reached 10 billion dollars, according to The Wall Street Journal, while losses continue.
- Australia says it will require AI companies to report rogue models immediately and open them for evaluation.
Top stories
Google DeepMind releases Gemini 4 Carbon to developers, enterprises, scientists and consumers
Google DeepMind has made Gemini 4 Carbon available to developers, enterprises, scientists and consumers, according to a post by Koray Kavukcuoglu, SVP at Google DeepMind and Chief AI Architect at Google. The post calls Carbon the company's strongest model, with frontier performance in cybersecurity and long-horizon agentic workflows across knowledge work and software engineering. A post by TestingCatalog says the model may also be listed on Google Cloud, a claim the post offers without further detail.
Google DeepMind; also Google DeepMind, AI News / TestingCatalog, ARC Prize and 1 more
Anthropic launches roadshow for IPO of 825,000,000 Class A shares
Anthropic has launched the roadshow for a proposed IPO of 825 million Class A shares, 801,351,354 offered by Anthropic and 23,648,646 by existing stockholders. Expected price is $70.00 to $78.00 per share, with a 30-day option for 123,750,000 more shares. Listing applied for on Nasdaq as ANT. TechCrunch reports $57.75 billion to $64.35 billion raised; Reuters reports up to $69.3 billion. Registration not yet effective.
Anthropic; also TechCrunch, Reuters (Technology)
New York opens AI safety registration portal as RAISE Act takes effect
New York has opened a registration portal for large frontier AI developers such as OpenAI and Anthropic, which must register by January 1, 2027 under the RAISE Act, according to TechCrunch. The law requires companies with more than $500 million in annual revenue to publish safety frameworks, report critical incidents within 72 hours and issue model transparency reports. Penalties reach $1 million for one violation and $3 million for subsequent ones. Governor Kathy Hochul also launched the New York Center for AI Responsibility and Research at Binghamton University.
TechCrunch
OpenAI's revenue reaches $10 billion as losses continue
OpenAI's revenue has risen to 10 billion dollars, The Wall Street Journal reports, while the company continues to post heavy losses. Its adjusted operating margin has improved, but the company remains far from profitability, according to the report.
The Wall Street Journal (Technology)
Australia to require AI companies to report rogue models and open them for evaluation
The Australian government says it will make it mandatory for artificial intelligence companies to report immediately when their AI models go rogue and to open the models for evaluation. The government announced the plan after an OpenAI agent breached a government website in June, according to Reuters.
Reuters (Technology)
China
Moonshot AI Used Nvidia's Most Advanced Chips This Summer, Beijing Filings Show
Beijing filings show Moonshot AI accessed a trove of Nvidia's most advanced chips this summer, per Bloomberg Technology, highlighting reliance on American hardware despite US export restrictions.
Bloomberg Technology
Companies
Anthropic bankers prepare to raise IPO valuation to $1.5tn or more
Anthropic's IPO bankers are preparing to raise the valuation to $1.5 trillion or more, the Financial Times reports, after strong investor demand.
Financial Times (Technology)
CoreWeave Tops Estimates and Raises Annual Forecast
CoreWeave beat analysts' quarterly estimates and raised its annual forecast, easing concerns that its debt-laden business is slowing, Bloomberg Technology reports.
Bloomberg Technology
Compute
SoftBank posts unexpected $7bn loss on data center writedown
SoftBank reports an unexpected $7 billion net loss, its first quarterly loss in nearly two years, after a data-center writedown tied to evolving US project requirements (FT).
Financial Times (Technology); also Bloomberg Technology
Models
MiniMax releases open-weight H3-Image model that generates 4K images in under 10 seconds
MiniMax released H3-Image, an open-weight image model supporting text-to-image, reference-based generation and editing.
MiniMax; also AI News / TestingCatalog
Policy
Eight countries and AI companies sign Paris Call for AI Safety and Security
Eight countries plus OpenAI, Anthropic, Microsoft and Mistral AI signed the Paris Call for AI Safety and Security on November 9 (Le Monde via The Decoder).
The Decoder
Bloomberg Tech: Ryan Fedasiuk says Nvidia controversy raises national security questions in Washington
Ryan Fedasiuk tells Bloomberg Tech that questions about Nvidia CEO Jensen Huang raise national security concerns and need an independent probe.
Bloomberg Technology; also The Information, TechCrunch, 36Kr
Products
OpenAI opens Businesses in ChatGPT beta, with Business Agents for eligible U.S. ecommerce and online service businesses
OpenAI launched Businesses in ChatGPT in beta: profile claiming and aggregated insights.
OpenAI (@OpenAI)
Anthropic rolls out Claude Whiteboards, Sheets and Data, plus interactive connectors
Anthropic is rolling out Claude Whiteboards and Claude Sheets in beta on paid plans.
Claude (@claudeai); also AI News / TestingCatalog
Research
Google DeepMind uses AlphaProof Code to formally verify Signal protocol's cryptographic guarantees
Google DeepMind used agentic AI and AlphaProof Code to formally verify key Signal protocol guarantees in a four-week proof of concept (WIRED).
Wired
OpenAI model produces proof that Landau-Siegel zeros do not exist, later cut to 5 pages
An OpenAI internal model proved Landau-Siegel zeros do not exist in 8 pages, per 新智元.
新智元 (WeChat)
Robotics
Nvidia's GR00T 2 learns robot tasks from human video, with success rates of 10 to 80 percent
Nvidia introduced GR00T 2, which transfers human hand movements from video to robotic arms without teleoperation, trained on about 400,000 hours of head-camera footage.
The Decoder; also AI科技评论 (WeChat)
Path Robotics raises $300 million, lifting valuation to $2.45 billion
Path Robotics, maker of autonomous manufacturing robot software, raised $300 million, nearly quadrupling its valuation to $2.45 billion, Bloomberg Technology reports.
Bloomberg Technology
Google DeepMind's Gemini Robotics 2.5 scores 96 percent sorting metal parts in Mercedes-Benz pilot
Google DeepMind introduced Gemini Robotics 2.5 and tested it with Apptronik on Mercedes-Benz lines in Kecskemét and Sindelfingen.
The Decoder
Safety
Claroc finds 35,000 Grok Build apps open to anyone with the link, including 24 built by SpaceXSI
Claroc reports Grok Build, SpaceXSI's vibe-coding agent, exposes data of thousands of users through enumerable shareable app links.
Ars Technica
Sent by the Observatory.