Category leaders

The Daily Debrief

Anthropic prices its IPO at $83 a share, Google's Gemini 4 Neon

Anthropic's IPO price, Google's Gemini 4 Neon, US auditors getting chatbots to plan attacks, Gemini's free Tasks agent and how long Google keeps your chats.

Issue 70 · Thursday, 19 November 2026

Anthropic priced its IPO at $83 a share, raising about $81bn That’s for 980 million shares, more than the 825 million Anthropic planned, at a price above the $70 to $78 range it set on 10 November. TechCrunch puts the valuation at about $1.5tn and says it’s the biggest listing ever by money raised, ahead of the $75bn SpaceX raised in June. The price was set on Wednesday, and trading starts on Nasdaq on Thursday under the ticker ANT. Source

Google’s Gemini 4 Neon costs $3 per million tokens in It sits between Carbon, Google’s model for the hardest, longest jobs, and Hydrogen, its fast one for everyday work, and Google says it costs 32% less per task than Carbon. It started rolling out on Wednesday for AI Pro and Ultra subscribers in the Gemini app, and on the API, where output costs $15 per million tokens. Source

US auditors got ChatGPT, Claude, Gemini and Grok to plan attacks Undercover testers at the Government Accountability Office got workable plans out of all four in under 12 hours, two for bombings and two for attacks on the power grid, getting round the safety blocks with common tricks. The report came out on Wednesday and doesn’t say which chatbot made which plan. Google and OpenAI said the tests used older models. Source

Google’s Gemini Tasks agent is now free in six countries It can phone businesses to book appointments and check out with Google Pay within limits you set, stopping to ask before anything like hitting Buy, though some of the extras start in the US only. It runs on Gemini 4 Hydrogen and started rolling out on Tuesday in the US, Canada, Australia, India, Japan and South Korea. It began in May as Gemini Spark, for a small group of testers. Source

Google will keep Gemini chats for two months with history off That’s up from 72 hours, and it covers temporary chats too. Google says it will use them only for safety and security, not to train its models, and it’s splitting chat history and AI training into two settings, so you can keep your history and still opt out of training. It made the change public on Monday, and it rolls out over the coming weeks. Source