The Maivia Gazette

Verified AI news, every morning

Thursday, November 5, 2026

Sources verified at 4:23 AM ET · complete

A long vault corridor of hundreds of small opened iron doors, each patched with brass, while a ceiling-mounted mechanical arm works its way along them.
AI-generated illustration, not event photography. The motion is AI-generated from the still.
Security

OpenAI says its automated pipeline built 948 working Linux root exploits in one run and wrote a patch for each

The run cost $1,213,430 in API credits, about $1,279 per exploit, and used an upcoming OpenAI model.

Verified · 3 sources Compiled from 3 verified sources · verified 4:23 AM ET

Film frames on a rooftop clothesline show a paper bird in successive flapping positions against a sunny skyline.
AI-generated illustration, not event photography. The motion is AI-generated from the still.
Models

OpenAI's Images 2.6 model adds animated GIFs to ChatGPT

Users can make GIFs from text or photos and edit them frame by frame. Developers get two new API image models.

2 verified sources

Security

Research

Models

Business & Policy

Sources in this edition

OpenAI says its automated pipeline built 948 working Linux root exploits in one run and wrote a patch for each

  1. RuntimeWireOpenAI says its agents exploited and patched 948 Linux kernel flaws for $1.2MPublished · fetched
  2. Help Net SecurityOpenAI’s upcoming AI model built 948 working Linux root exploits in a single run - Help Net SecurityPublished · fetched
  3. The Hacker NewsOpenAI AI Pipeline Generates 948 Linux Root Escalation Exploits in 30-Hour RunPublished · fetched

Zyxel patches a GS1900 switch zero-day that was used to compromise nearly 1,000 devices

  1. BleepingComputerZyxel fixes zero-day exploited to hack nearly 1,000 network switchesPublished · fetched
  2. The Hacker NewsZyxel GS1900 Patch Fails to Block Zero-Day Exploit Allowing Root Command ExecutionPublished · fetched

Bitdefender's GhostMap bug lets a permission-free Android app trap phones in a boot loop, and Google won't issue a security fix

  1. BleepingComputerGhostMap Android flaw lets malicious app trigger persistent boot loopPublished · fetched
  2. Help Net SecurityA banking malware sample led researchers to an Android bug that can loop-crash phones - Help Net SecurityPublished · fetched

CloudOpsBench finds AI agents fail many real AWS operations tasks, usually because of reasoning mistakes

  1. The DecoderNew open-source benchmark reveals reasoning as the biggest bottleneck for AI in cloud operationsPublished · fetched

MIT researchers show GPT-6 Astra can read and edit other AI models' weights directly, without retraining

  1. The DecoderGPT-6 Astra sees the matrix and reprograms other AI models directly through their weightsPublished · fetched

Moonshot AI publishes the weights of its 2.8-trillion-parameter Kimi K3.1 Pro model

  1. RuntimeWireMoonshot AI releases Kimi K3.1 Pro weights for its 2.8-trillion-parameter modelPublished · fetched

OpenAI's Images 2.6 model adds animated GIFs to ChatGPT

  1. TNW | AppsChatGPT can now make animated GIFs with OpenAI's Images 2.6Published · fetched
  2. The VergeChatGPT will now make you animated GIFsPublished · fetched

Voters reject data centers in all but one ballot measure as Democrats win control of Congress

  1. CNBCVoters rebuff data centers in every ballot measure except one during the 2026 midtermsPublished · fetched
  2. TechCrunchAfter the midterms, a bigger clash over data centers could be coming | TechCrunchPublished · fetched

Covers 4:23 AM Nov 4 – 4:23 AM Nov 5 ET · generated · Permanent link to this edition