AI news, 4 November 2026

Google disables Search for K‑12 schools that reject its AI features

Google will turn off Search and Gemini services for K‑12 student accounts in schools that opt out of its AI, with administrators unable to disable Gemini without also disabling Search, per a statement to The New York Times and a broader education update.

Toyota unveils 16nm SoC with STT‑MRAM that boots in 65.7 ms

Toyota, via its AISIN partner, built a 16nm FinFET SoC with an Arm Cortex‑A53 dual‑core CPU, Mali‑G31 GPU and integrated STT‑MRAM that stores model weights and the OS, achieving a 65.7 ms power‑on to kernel startup—38× faster than using external memory—and running Tiny YOLOv3.

Soket Labs and ZenteiQ launch EKA-1 and Dhīr models under IndiaAI Mission

Soket Labs unveiled the 26.3 B‑parameter EKA-1 model, trained on 7.5 T tokens and supporting 22 Indian languages, while ZenteiQ released Dhīr 2B and Dhīr 5B multilingual assistants for low‑connectivity environments.

Qwen 3.8 27B runs weight‑space programming on laptop

A blog‑style paper shows GPT‑6 Astra can read and edit raw model weights; the author demonstrated Qwen 3.8 27B on a laptop performing the same weight‑space tasks, improving MATH‑500 from 7.9 % after editing 0.016% of weights.

SETI Institute and Breakthrough Listen Deploy AI Agents to Control Allen Telescope Array

The SETI Institute and Breakthrough Listen are using Nvidia’s Holoscan platform to run AI agents that steer the Allen Telescope Array and analyze radio data in real time. The system reduced tasks that previously took weeks to hours and detected a Crab Nebula pulsar live; Nvidia claims the hardware can process up to 2.4 terabits per second.

Google unveils Gemini 4 Carbon, beats Claude Opus 5.5

OpenAI uses GPT-6 Astra to design hardware device in under a week

Google launches Antigravity agent with Cheminformatics Skills to speed drug discovery

Anthropic Introduces Message Threads, Cutting First‑Token Latency 35% for Long Chats

Anthropic blocks Claude Code from cloning existing video games

HCLTech launches AI agents and conversational assistants for mainframe system administration

OpenAI launches Pando backend, cutting median Git push latency by ~80%

Anthropic releases FlashProve zkSNARK prototype proving model replies

Hugging Face launches StateBench benchmark for long‑context state tracking

Collinear AI launches CWE-bench v2 with 100 public security tasks

MIT researchers use LLMs to edit neural network weights, boosting MATH-500 score

Stanford study finds 44% of LLM claims expire within 15.7 months

Hugging Face Papers highlight agentic LLMs generating and refining graphics code

OpenAI says ChatGPT handles 20% of Google’s search query volume

Apple adds Siri Recap and Live Rewind to watchOS 27.3 beta

Netflix launches AI Discover tab with themed recommendation experiments

Palette fine‑tunes 20B model to generate design‑language slide decks

AMD earnings lift stock 8% as data center revenue nearly doubles

Revolut cuts AI coding costs 90% by self‑hosting Chinese GLM model

US tech giants' hidden AI debt rises 90% to $3.2 trillion in six months

Fal Raises $1.3B, Valuation Doubles to $16B Amid Inference Demand Surge

Anthropic Compute Costs Rise to 58% of $17B Revenue, Margins Worsen

CrowdStrike Said to Pay $495M for AI Security Firms Fiddler and Dreadnode

Brookfield announces $27 bn AI data center project at UK power station

NVIDIA outlines AI tools accelerating semiconductor fab simulations

Drax partners with Brookfield and Radiant to plan £20bn AI datacenter, 100MW phase by 2028

Huawei directs customers to source HBM3E memory on black market as chip supply tightens

NVIDIA expands physical and agentic AI platforms for cities, enabling AI services on RTX workstations

Binance reports $6.43 B theft via fake job interviews and tampered Homebrew AI tool

OpsShield’s cPGuard update channel compromised, backdoor deployed to versions 5.88.12‑5.88.19

Google reports LLMjacking fraud generates up to $1 M daily cloud AI charges

European supermarkets’ fridges remotely heated in cyberattack

BTCPay Server urges upgrade after Bitcoin and Litecoin thefts from versions ≤2.4.1

Synology urges DSM update after ransomware exploits Dec 2024 patched flaw

Apple warns AI agents could exceed permissions or resist instructions in annual filing

UK Cyber Chief warns AI cyberattacks on critical infrastructure by 2028

E2E Networks launches Blaze, a pay‑per‑token AI inference service

TRAI permits AI agents for telecom support, mandates 16‑hour human assistance

Dispatch Network raises ₹5 Cr pre‑seed from Antler India to expand AI team

Delhi Police deploy facial‑recognition vans to South Delhi mall, sparking surveillance concerns