AI news, 31 October 2026

Google AI Mode cuts Search revenue 15% ($500 B), worst since 2004 IPO

Google reports its AI Mode feature reduced Search revenue by 15%, a $500 B decline, the steepest drop since its 2004 IPO.

Intel stock falls 28% after Q4 outlook miss, wiping $12 bn from government stake

Intel’s shares dropped about 28% after the company forecast Q4 revenue of $15‑16 billion, below analysts’ $17 billion expectations, and said gross margin would fall to 37% from 43%. The decline cut the value of the government’s 10% stake by roughly $12 billion.

Google satellite boots TPUs, runs Gemma 4 model in orbit

Google reported that its AI‑chip satellite successfully booted its Tensor Processing Units and ran the on‑board Gemma 4 model to generate a space‑themed haiku, part of Project Suncatcher after launch earlier this month.

Microsoft Research and UC Berkeley launch TokenEconomizer plugin cutting coding agent token costs up to 30.9%

TokenEconomizer, an open‑source plugin from Microsoft Research and UC Berkeley, learns from past sessions and reduces coding agent token usage by 18.4‑30.9 % on SWE‑Bench without altering the agent.

Zest launches AI restaurant concierge that books reservations via text

Zest introduced a personal AI‑powered restaurant concierge that lives in text messages, combines users’ tastes, lets friends vote on options, and books tables directly, leveraging credit‑card data to track actual dining history. Founder Mario Gomez‑Hall says the agent is now the product.

OpenAI publishes fair‑use brief excerpts

Drew Breunig embeds NES emulator in Qwen3.5-4B weights

Contractor links Claude Agent to Pentagon system, Anthropic files IPO

Cisco gives engineer unlimited AI tokens to rewrite legacy code

Anthropic, OpenAI, Cloudflare, Google release AI agents that hunt code security bugs

Samaya AI unveils Sterling 3B finance research model beating Claude Fable 5.1 with 11× lower cost

Creator builds MonsterHouse Halloween projection using NVIDIA Nemotron Voice Agent and related models

Edge0 launches Audio8-TTS-Realistic 0.6B open-weight voice cloning model

Researchers propose Distribution Matching Distillation for text-to-image generation

Authors introduce Adaptive Task Refresh for zero-data multimodal self-improvement

Researchers argue benchmarks encode difficulty distributions and identify measure mismatch and coverage gaps

Youtu-Parsing-Omni Introduces Unified Encoder for Documents, Images, Audio, Video

Look Only When Needed Proposes Certificate‑Guided Visual Routing for Diffusion VLMs

Researchers Prove Platonic Representation Theorem for Joint‑Embedding Predictive Architectures

Apple launches Apple Intelligence in China via iOS 27.3 beta

Spotify plans AI remix service launch in early 2027

Ollie AI assistant shuts down on November 6

Figma launches Weave, AI generation and editing workflow tool

OpenAI raises under half of $30 B target

Anthropic likely to receive double‑B junk rating after IPO per FT

Nvidia talks to co‑sign billions of debt for Nscale’s West Virginia AI data center

Private equity adds AI disruption clauses to buyout loan agreements

Anthropic's Claude Code Extension Trojanized on Cursor Marketplace with Persistent SSH Backdoor

Chinese Hackers Deploy AI‑Powered ARTEX Framework Against Over 200 Organizations

Trezor alerts users to suspected implant in Safe 5 hardware wallets

WordPress core RCE flaw triggers ~63,000 attacks within 24 hours

OpenAI, Google, Microsoft, Anthropic Adopt Policy Token to Let Publishers Opt Out of AI Scraping

Wikimedia Foundation Warns AI Copyright Ruling Could Threaten Open Web

OpenAI Revises Model Spec, Adding Memory Use and Assumption Disclosure Requirements

Prateek Jogani launches AI-native consultancy with $6.6M backing

MeitY proposes DEPA framework to share AI training data with privacy safeguards