Anthropic sets its IPO price range
Anthropic began its IPO roadshow on 10 November (US time). It's selling 825 million Class A shares at $70 to $78 each, and they'll trade on the Nasdaq Global Select Market as "ANT". Of those shares, 801,351,354 are new stock from Anthropic. Existing holders are selling the other 23,648,646, and Anthropic gets no money from that portion. Underwriters can buy up to 123.75 million more shares within 30 days. The registration statement is filed with the SEC but hasn't become effective yet.
CNBC puts the top of the range at about $64.35B raised and a fully diluted value of up to $1.405T. That's up from $965B in Anthropic's May funding round. CNBC also reports that AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC will buy $26B of stock at the IPO price in a separate private placement. It says pricing is planned for 18 November and the first trades for 19 November.
For builders: Once Anthropic is listed, its pricing and product choices will answer to quarterly results. If your product depends on Claude, keep a second provider tested and ready.
Gemini 4 Carbon is generally available
Google released Gemini 4 Carbon on 10 November, two weeks after a limited release to cyber defenders through its Fairwind Program. Google calls it its strongest model and built it for complex multi-step coding, open-ended science and long agent runs. Paying developers can use it through the Gemini API in AI Studio and in Android Studio, and enterprises get it through Gemini for business. Google AI Ultra subscribers get it in Antigravity, the Gemini app and AI Mode in Search.
The launch price matches Argon at $2 per million input tokens and $10 per million output tokens. After 30 November it goes up to $4 and $20. On Google's own numbers, Carbon ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench. Against an adaptive red-teaming agent, its per-attempt breach rate fell to 0.8%, down from 3.1% for Argon. Google still recommends Argon for most software engineering and knowledge work, because Carbon spends more tokens on reasoning and checking its own work.
For builders: Send your hardest agent and debugging jobs to Carbon and keep routine traffic on Argon. Base your budgets on the December rate, which is double the launch price.
Claude gets Sheets, Whiteboards and Claude Data
Also on 10 November, Anthropic added three new artifact types to Claude. Claude Sheets is a shared spreadsheet with formulas and multiple tabs that people and Claude build together. Claude Whiteboards is for laying out ideas in meetings, and Claude can summarise the result. Claude Data, which used to be called Dashboards, pulls from tools such as Snowflake and Salesforce to make dashboards and short narrated videos. Claude Meetings, which will join calls to take notes, is listed as coming soon. Interactive connectors now let partners such as Adobe and Atlassian run their apps inside Claude, and the marketplace lists more than 9,000 connectors and plugins.
For builders: If you make Claude connectors, you can now put a working app view inside the chat. Users don't have to follow a link out to your site.
OpenAI opens ChatGPT to businesses
On 10 November, OpenAI launched a beta called Businesses in ChatGPT. Verified companies get a profile with a blue check in the Business Hub. They can update their details, prices and stock there, and see how often ChatGPT mentions them, how many users click through and what topics come up. Eligible US ecommerce and online-service companies can also run Business Agents. These answer customers in a labelled chat that's kept apart from ChatGPT's own answers, and businesses can sponsor them as the landing spot for ChatGPT Ads.
For builders: If your team sells online, claim the profile and keep your product data current, because ChatGPT's agent answers from that data.
Australia plans an AI safety regulator
On 10 November, Prime Minister Anthony Albanese said AI companies will face mandatory safeguards, enforced by a new AI safety regulator. Companies would have to report AI security breaches, and frontier models would be evaluated independently. The move follows an incident in June, when a prototype OpenAI model got into part of a government health statistics portal. OpenAI didn't tell the government until September, and it did so by emailing a general inbox that's checked once a day.
For builders: If your agents touch live systems in Australia, start logging what they do and write down an incident process now.
DeepMind releases AI-written proofs for Signal's code
Google DeepMind has published two AI-generated Lean proofs of security for the official Rust libsignal library (v0.105.4), checked against the compiled AArch64 machine code. One proof uses a symbolic attacker model and the other a game-based computational model. The work was covered on 11 November and is released under Apache 2.0. The scope is narrow. The proofs target a command-line test harness with the cryptographic primitives hooked out, and they don't cover the Signal apps people install.
For builders: AI tools can now produce machine-checked proofs for real security code. Read the stated assumptions before relying on any result like this.
Railway reports $740,000 in fraudulent AI charges
On 10 November, Railway published a new report on the breach that began on 20 October. Attackers broke out of its build system's isolation and then used a misconfigured trust relationship to get credentials that decrypt customer environment variables. They went after AI provider keys and ran about $740,000 of unauthorised usage. At most 1.85% of workspaces, about 100,000, had their variables accessed. Railway says it'll pay any charges that AI providers won't refund.
For builders: Treat AI API keys like payment cards. Set spend caps and alerts with every provider, and rotate any key stored on a platform that's been breached.
Sources
- Anthropic: Roadshow launch
- Google: Gemini 4 Carbon now available
- Claude: Sheets, Whiteboards and interactive connectors
- Google DeepMind: code-verification repository
- CNBC: Anthropic sets price range of $70 to $78 in upcoming IPO
- The Next Web: OpenAI Businesses in ChatGPT and Business Agents
- TechXplore: Australia to impose mandatory AI safeguards
- BleepingComputer: Railway breach leads to $740,000 in fraudulent AI charges
- The Neuron: Everything that happened in AI today, 11 November 2026
Talk about it with people building with AI
Our Auckland meetups are open to everyone. Members get a ForgeBox and the members' projects area for NZ$599 a year.