AI news, 1 November 2026

Cursor Marketplace Distributes Malicious Claude Code Extension to 5,000 Developers

A malicious version of Anthropic's Claude Code VS Code extension (version 2.1.313) was available in Cursor's extension marketplace for about 11 hours on Oct 29, exposing up to 5,000 developers to malware that installs persistent SSH backdoors.

OpenAI raises $11 B, under half of $30 B target, plans $20 B family‑office round

OpenAI secured about $11 billion in its latest round, valuing the company at $1.25 trillion, well below its $1.4 trillion goal. It will seek up to $20 billion from family offices and plans an IPO early next year, while rival Anthropic eyes a November IPO.

Anthropic's Opus 5.5 blocks video game cloning and reverse‑engineering

Anthropic's Opus 5.5 model blocks attempts to recreate video games and reports indicate it also blocks non‑malicious generic reverse‑engineering tasks, prompting complaints from game developers.

Anthropic Exploits Critical Linux Kernel Flaw to Gain Remote Root on MPTCP Servers

Anthropic researchers developed an exploit for CVE-2026-98432, a use‑after‑free in the kernel's IPv4 TCP code with a 9.8 severity score, affecting Linux kernels from version 4.4 including Ubuntu, RHEL and Rocky Linux. The exploit enables unauthenticated remote code execution as root on servers running Multipath TCP listeners, though RCE on plain TCP has not been publicly demonstrated.

Meta reports Muse reaches over 10 M users, 25% shop via AI assistant

Meta says more than 10 million users in the US and Canada have activated Muse, with over 25 % using it for shopping and an average purchase size above $200. Executives at roughly 50 companies said they expect to benefit rather than be disrupted.

OpenAI will honor shared robots.txt signal for AI training preferences

OpenAI adds expanded violence safeguards and teen protections to ChatGPT

Google launches first data‑center satellite running Gemma 4 model

Arca builds internal memory system and custom model harness for ten‑day financial agents

Datasette‑LLM plugin now auto‑detects default model, simplifying agent setup

Factory launches Droid 1.0 coding agent harness with model‑independent tool calling

Vercel Labs open‑sources agent‑desktop CLI for AI agent computer use

GitHub repo jefferysha0627/agent-office-3d showcases live AI agent status

Memdoor team of five Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard

Exin Therapeutics uses AI agent to inject gene therapy into live mouse brain

Claude Sonnet 5 autonomously selects gene therapy and performs mouse brain surgery

L3G-World Enables Real-Time Infinite World Generation with Latent 3D Gaussian Memory

SpecSlot Introduces Role‑Coupled Tokenization for Efficient Pixel‑Space Generative Transformers

WorkFactory Framework Generates Executable Professional Tasks for Scalable Agent Training

Starbucks to test AI‑driven personalized drive‑thru menus starting 2027

Meta’s Muse Charm AI Tamagotchi sells for $50, built on open‑source Muse

Darkbloom warns low GPU rental payouts as token prices plunge

Google adds more ads to AI search as AI‑search revenue slows

Microsoft plans to raise $25 billion in debt for AI buildout

Runware aims to mass‑produce AI data center pods at 30 MW per day by 2028

Boom Supersonic shifts to AI data center turbines, airliner timeline in doubt

Fujifilm invests $63M to quintuple LTO tape production for AI archiving

Parameter Labs Exploit Lets Attackers Take Over ChatGPT Accounts via HTML Previews

DOJ Cracks Down on AI‑Powered ‘Ghost Student’ Scheme, Seizes $750K in Crypto

Anthropic expands watermark to four Claude models, leaves bypass on Opus 4.6 and Sonnet 4.6

OpenAI will train on Daybreak Blue session data from Nov 29, adds opt‑out after backlash

Amazon issues internal memo on AI use, sparking employee unease

MeitY launches DEPA handbook to boost trusted AI data sharing