# AI news brief: Saturday, October 31, 2026

7 stories, most important first. Status is how solid each one is: official, confirmed, reported or rumor.

- [Google DeepMind introduces Gemini 4 Carbon, aimed first at cyber defenders](https://ai-news.ikangai.com/story/google-deepmind-introduces-gemini-4-carbon-aimed-first-at-cyber-defenders/index.md): reported. Google DeepMind's Gemini 4 Carbon targets cybersecurity and long-running engineering work, with access starting for a subset of cyber defenders.
- [Parameter Labs details ChatGPT HTML preview flaw that could expose user data](https://ai-news.ikangai.com/story/parameter-labs-details-chatgpt-html-preview-flaw-that-could-expose-user-data/index.md): reported. Parameter Labs describes an attack chain in ChatGPT's HTML previews that let a malicious share link reach other signed-in users' private data, which OpenAI has since closed.
- [Claude adds invisible text watermarks, with detection limited to selected parties](https://ai-news.ikangai.com/story/claude-adds-invisible-text-watermarks-with-detection-limited-to-selected-parties/index.md): reported. A Reddit post citing Anthropic's help page says Claude's statistical text watermark is extended globally to several models from November 3, and detection is restricted.
- [Anthropic says malicious Claude Code extension in Cursor marketplace was not its build](https://ai-news.ikangai.com/story/anthropic-says-malicious-claude-code-extension-in-cursor-marketplace-was-not-its/index.md): reported. A malicious copy of the Claude Code extension in Cursor's marketplace reached at most 5,000 users, and Anthropic says it did not build or publish it.
- [Arca says its agents run tasks up to ten days, with humans approving trades and payments](https://ai-news.ikangai.com/story/arca-says-its-agents-run-tasks-up-to-ten-days-with-humans-approving-trades-and-p/index.md): reported. Arca's head of AI says its agents handle long-running wealth management work using a custom harness and memory system, while humans make final financial decisions.
- [Microsoft's 4B hybrid computer-use agent scores 66.0% on OSWorld-Verified](https://ai-news.ikangai.com/story/microsoft-s-4b-hybrid-computer-use-agent-scores-66-0-on-osworld-verified/index.md): reported. A Microsoft 4B hybrid agent that can both click and write code scores 66.0% on OSWorld-Verified, ahead of 27B and 30B agents, per a post.
- [OpenAI is building Private Intelligence, an enterprise tier with stronger data controls](https://ai-news.ikangai.com/story/openai-is-building-private-intelligence-an-enterprise-tier-with-stronger-data-co/index.md): reported. Private Intelligence, an enterprise offering with stronger data controls for businesses using OpenAI's top models, is being built with large customers.

---

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it.
