News
This Week in AI: October 26, 2026
Anthropic's IPO filing, Texas suing OpenAI, Claude Duet, OpenAI on Google Cloud, Kimi K3.1 Pro, and the UK's AI biosecurity lab — this week's AI news.
Anthropic filed to go public on Monday, capping a week in which AI's biggest names kept colliding with courts, clouds, and each other. Texas became the tenth state to sue OpenAI, Anthropic launched an agent that works while you are away, and OpenAI's models landed on rival Google's cloud. Here is what mattered, in plain English.
Anthropic files for its IPO
Anthropic filed its long-awaited IPO prospectus on Monday, October 26, listing on the Nasdaq as ANT with Goldman Sachs and Morgan Stanley as lead underwriters, according to CoinDesk's live coverage. It disclosed first-half revenue of $16.3 billion, versus $1.2 billion a year earlier; the amount to be raised was not yet disclosed. Meta announced a $95 billion share sale the same morning.
Why it matters: An Anthropic IPO would put one of the two leading US AI labs under public-market scrutiny for the first time, with quarterly numbers to defend. Takeaway: The AI race is now a capital-markets story, not just a technology one.
Texas sues OpenAI — the tenth state
Texas Attorney General Ken Paxton sued OpenAI and CEO Sam Altman on Friday, October 23, in Collin County district court, making Texas the tenth state to sue the company. The petition, brought under the Texas Deceptive Trade Practices Act and the SCOPE Act (the state's minors' online-safety law), alleges ChatGPT was marketed as safe while failing to protect users, including children. Texas asks the court to bar ChatGPT for minors in the state and require a court-approved evaluator to review new models before Texans can use them, plus $10,000 per violation in penalties. Florida sued in June and eight more states filed on October 14; only Texas and Florida name Altman personally. These are allegations, not court findings.
Why it matters: If a judge ever granted the model-vetting request, a single state court would effectively hold a veto over frontier model releases in the second-largest US state. Takeaway: The fight over AI safety is moving from Congress to state courtrooms.
Claude Duet: an agent that works while you are away
Anthropic opened a waitlist on October 21 for Claude Duet, a proactive agent app built on Claude Opus 5.5, in a research preview for US Pro and Max subscribers. Unlike a chatbot that waits for your question, Duet is designed to flag what matters and get work done before you ask — the latest step in the industry's shift from chatbots you prompt to agents you delegate to. Anthropic also put the full Claude Platform on Google Cloud and donated its Agent Skills standard to the Linux Foundation.
Why it matters: Agents that act unsupervised raise the stakes on reliability and permission: useful when they are right, costly when they are not. Takeaway: Try delegated agents on low-stakes tasks first, and check their work.
OpenAI's models arrive on Google Cloud
OpenAI announced on October 21 that its frontier models and Codex are coming to Google Cloud's Model Garden, payable against customers' existing cloud commitments. On October 22 its Decisions API reached general availability at $0.10 per million input tokens, with no output-token charge. Also on October 21, OpenAI updated its usage policies: from November 20, model routers and AI gateways giving third parties direct access to its models must get its review and approval first.
Why it matters: The fiercest AI rivals are now also each other's distributors, and the cloud giants are happy to sell everybody's models. Takeaway: For businesses, model choice is increasingly a procurement decision, not a loyalty test.
Kimi K3.1 Pro and MiniMax M3.1 Flash keep the open-weight wave rolling
Moonshot AI released Kimi K3.1 Pro on October 23, a post-trained successor to Kimi K3 with a one-million-token context window, plus a Highspeed variant; full weights are promised by November 4. Independent benchmarks rank it near the top of public models for agentic and cybersecurity tasks. MiniMax's M3.1 Flash, launched October 19, costs $0.15/$0.60 per million input/output tokens with the same window.
Why it matters: Strong open models keep downward pressure on prices and give smaller developers frontier-class tools without a big vendor contract. Takeaway: The cheapest capable model for your task keeps changing — re-check pricing before you lock in.
The UK bets £100 million on AI biosecurity
The UK government announced on October 24 a new AI Biosecurity Lab: £100 million in public money plus up to £200 million in model credits, compute, and support from OpenAI and Anthropic, with Google DeepMind and Basecamp Research as research partners. It will use frontier models to spot biological threats earlier and speed up vaccines and treatments. Only the £100 million is cash; the rest is in-kind.
Why it matters: Biosecurity is one of the few AI-risk areas where governments, labs, and researchers broadly agree on acting early. Takeaway: Expect "AI for defense" — against disease, not just cyberattacks — to be a growing public-spending theme.
Also this week: Block unveiled Dessa, a voice model that can listen and speak at once; Google added a Gemini hub for small businesses; ChatGPT Ads expanded to 89 countries; and Canva launched its first image model, Craft.
Markets & crypto this week
It was a choppy week, and AI stocks were at the center of it. The Nasdaq jumped 1.1% on Wednesday on tech earnings, then plunged 2.6% on Thursday — its worst session since June — on Meta's AI spending plans and a gloomy warning from chip-equipment maker Lam Research. For the week, the S&P 500 fell 0.47%, the Nasdaq 100 fell 0.79%, the Dow gained 0.41%, and the Russell 2000 rose 2.0%. Bitcoin opened near $82,000, peaked around $84,000 on Wednesday, then slid all week and through the weekend, dipping below $80,000 on Saturday before recovering to about $80,800 on Sunday — roughly 2.4% lower. Ether drifted from about $2,423 to near $2,400. Winners included Super Micro (up 9.6%), Spotify (up 9%+), and Crown Castle (up 12%). Drivers: AI spending worries, spiking oil, about $773 million of Bitcoin ETF outflows on Thursday and Friday, and de-risking ahead of the Fed's October 27–28 meeting. Daily briefings: https://blutoolboxhq.com/markets.html.
Not financial advice — prices move fast; this is a recap, not a recommendation.
Money & the world this week
Oil was the week's wallet story. West Texas Intermediate crude settled 3.82% higher at $103.27 on Friday and Brent rose to $116.72, after reports the US and Israel may be drawing up plans for renewed attacks on Iran, following Houthi drone strikes on Saudi Aramco facilities. The 10-year Treasury yield hit 5.00% on hot economic data — a level that tends to push up mortgage and loan rates. By Monday, Brent had eased below $110 on hopes of an Iran deal. For households: higher crude lifts gas and heating costs, and higher yields keep borrowing expensive. Traders now watch the Fed's October 27–28 meeting for a December rate-hike signal. Background: https://blutoolboxhq.com/world.html.
Not financial advice — context to help you understand the economy, not a recommendation.
What to watch next week
- The Federal Reserve's October 27–28 meeting and any signal on a December rate move.
- Whether Anthropic's IPO filing reveals pricing and valuation details — and how markets react.
- Whether any court acts on Texas's request to restrict ChatGPT for minors or vet new models.