AI News Today: Anthropic Files IPO (Oct 29, 2026)

Anthropic heads for Nasdaq

The biggest story this week is Anthropic's public S-1 filing with the SEC on October 26, confirming plans to list on Nasdaq under the ticker ANT. The filing discloses $16.3 billion in first-half 2026 revenue against an $80.7 billion net loss, underwritten by Goldman Sachs and Morgan Stanley. For users, this matters because it shows how much capital AI labs are burning to build frontier models — and it gives the public its first detailed look at Anthropic's finances, which could shape investor and regulator expectations for the whole industry.

Mistral opens up its biggest model yet

French lab Mistral AI released open weights for Large 4, nicknamed "Le Chonk," a 1.05-trillion-parameter mixture-of-experts model with roughly 49–52 billion active parameters at any one time. It was trained on 4,000 Nvidia Grace Blackwell GPUs after an API preview on October 6. Open weights mean developers and researchers can download, inspect, and fine-tune the model themselves rather than relying solely on a paid API — a meaningful boost for open-source AI in Europe.

Infrastructure race continues

At GTC Berlin, Nvidia and Google Cloud announced general availability of Nvidia's next-generation Vera Rubin Infrastructure on October 23, plus new "DSX POD" AI factory systems meant to scale up Europe's AI compute capacity. Meanwhile, Amazon open-sourced TorchNeuron at the PyTorch Conference on October 20, letting developers run existing PyTorch code on AWS's custom Trainium chips without rewriting it — part of AWS's effort to offer a cheaper alternative to Nvidia GPUs. Together these moves signal that the fight over who supplies AI computing power is intensifying, which could eventually lower costs for developers.

Products: Copilot gets a makeover, Grok gets a voice

Microsoft shipped a major Copilot overhaul on October 23, adding a Home hub, a Code Builder agent, an Autopilot agent system, and a friendlier assistant persona called Mico. The goal is to make Copilot feel like a daily companion rather than just a chatbot. Separately, xAI launched Grok Voice 3.0 on October 14, a speech-to-speech API aimed at conversational and customer-support use cases, continuing xAI's push to position Grok as an "agentic operator." For everyday users, these updates mean more natural voice interactions and more autonomous AI agents handling multi-step tasks.

Regulation: EU AI Act gets its first technical standard

The European Commission referenced the first harmonised standard under the AI Act — EN 18286:2026, covering AI quality management systems — in an Implementing Decision dated October 12 and announced October 19. This is a concrete step toward operationalizing compliance ahead of the high-risk AI deadlines, which have been pushed to December 2027 and August 2028. For developers building AI products for the EU market, this standard gives the first real technical benchmark to work toward, rather than just legal text.

What it means

This week's news shows three parallel tracks: the business side (Anthropic's IPO filing), the technical side (bigger open models and faster chips), and the compliance side (the EU's first concrete AI Act standard). Developers should watch Mistral's open weights and AWS's TorchNeuron release closely, since both lower barriers to building and running large models outside the biggest proprietary platforms.