// AI Briefing · October 31, 2026

AI Briefing

Today’s video is on its way

The stories are ready to read now. The video usually follows within the hour.

Rendering the short

The iShares Semiconductor ETF ended the week 29% below its recent high, which puts it in a bear market. Mistral released open weights for Mistral Large 4 ('Le Chonk'), a natively multimodal mixture-of-experts model with about 1 trillion parameters, on Hugging Face and its API. A malicious version 2.1.313 of the Claude Code extension, published under Anthropic's legitimate extension ID, was distributed through Cursor's marketplace on October 29. CrowdStrike says the China-nexus group Vapor Panda used ARTEX, an agentic intrusion framework that runs mainly on DeepSeek, against more than 200 organizations in at least 14 countries starting October 17. OpenAI, Google, Microsoft and Anthropic say they will honor a single robots.txt signal that lets websites opt out of AI crawling. Google's Project Suncatcher prototype, launched this month on a Falcon 9, is now operating in orbit with four TPUs running DeepMind's Gemma 4.

AI Stocks Slide Into a Bear Market as Investors Demand Profits From AI Spending

The iShares Semiconductor ETF ended the week 29% below its recent high, which puts it in a bear market. On Friday Intel fell 25% after weak guidance, Micron dropped 5.5% and Nvidia 4%. Google had its worst trading day ever on Thursday after its core ad sales fell, and executives said AI Mode queries currently earn less than regular search. The wider market is holding up, so this is a targeted vote of doubt in the AI build-out: investors now want proof of returns before they back more capex.

02Open SourceTechCrunch ↗

Mistral Releases the Weights of Its Trillion-Parameter Large 4 Model, Pitched at Cyber Defense

Mistral released open weights for Mistral Large 4 ('Le Chonk'), a natively multimodal mixture-of-experts model with about 1 trillion parameters, on Hugging Face and its API. Mistral says it scores 51% on the Artificial Analysis Cyber Index and ties China's GLM-5.3 on the Coding Agent Index while using 47% fewer tokens per task. It is the biggest open-weight release yet from a Western lab. Mistral is selling it as 'sovereign' cyber defense that governments and businesses can run themselves. A penetration-testing variant is limited to vetted partners. The license requires model-as-a-service providers to buy a separate commercial license.

03IndustryThe Hacker News ↗

Trojanized Claude Code Extension on Cursor's Marketplace Planted an SSH Backdoor

A malicious version 2.1.313 of the Claude Code extension, published under Anthropic's legitimate extension ID, was distributed through Cursor's marketplace on October 29. Cursor says up to 5,000 users installed it. The loader added an attacker's SSH key, tried to enable SSH, opened a reverse tunnel, and sent host details back to the attacker. Cursor says the build was not made by Anthropic, and there is no evidence it reached Microsoft's marketplace or Open VSX. How the malicious update got published is still unclear. The incident shows that AI coding tools, which run with deep access to developer machines, have become a target for supply-chain attacks.

04ResearchCrowdStrike ↗

CrowdStrike: China-Linked Hackers Used a DeepSeek-Powered Agent Framework Against 200+ Organizations

CrowdStrike says the China-nexus group Vapor Panda used ARTEX, an agentic intrusion framework that runs mainly on DeepSeek, against more than 200 organizations in at least 14 countries starting October 17. The company confirmed post-exploitation activity at nine of them. Planner and worker agents carried intrusions from initial access through lateral movement, with a human operator steering at key decisions. CrowdStrike says this is the group's first observed use of AI-driven tooling. It is a documented case of a state-linked actor scaling espionage with an open Chinese model.

05PolicySiliconANGLE ↗

OpenAI, Google, Microsoft and Anthropic Back One Shared Opt-Out for AI Scraping

OpenAI, Google, Microsoft and Anthropic say they will honor a single robots.txt signal that lets websites opt out of AI crawling. A draft of the standard has been submitted to the IETF. Today publishers must block each AI company's crawlers one by one, so a single signal greatly simplifies opting out. OpenAI and Microsoft plan support in the coming months and Google next year. Meta backs the idea but hasn't committed to implementing it.

06ResearchTechCrunch ↗

Google's First Data Center Satellite Is Online and Running Gemma 4 in Orbit

Google's Project Suncatcher prototype, launched this month on a Falcon 9, is now operating in orbit with four TPUs running DeepMind's Gemma 4. Power limits mean only one TPU can run at a time, for about 20 minutes per run. Google will test how the chips handle radiation and heat. If they hold up, two more satellites with more power and optical links between them are planned for 2027. It is the first real test of whether AI compute can move to space as power on Earth runs short.