AI news — 8 November 2026

Every morning we collect the most important AI news from dozens of sources and sum it up in a few lines, with the newest open-source projects beside it. Each card links to the original source.

RSS · Today's newsRSS · New repositories

News 8 stories

The Decoder
Research

Open-Weight AI Models Narrow the Cyber Capability Gap with Proprietary Systems

Since open models cannot be restricted once published, defenders face a narrowing window to prevent breaches. However, organizations can also leverage these identical systems to detect and patch weaknesses.

Original

The gap between open and closed AI models in cyber capabilities keeps shrinking

In the Vals AI benchmark, OpenAI's GPT-6 Sol Daybreak hits 80.8 percent of attack objectives in simulated corporate networks. Worse, the open-weight model DeepSeek V4.1 Flash reaches 72.8 percent at a tenth of the cost. Because open models can no longer be restricted, analysts say the window for defenders is shrinking fast. The article The gap between open and closed AI models in cyber…

Read at the source Also covered by: WIRED
The Decoder
Models & releases

Nous Research fine-tunes Alibaba's Qwen model to strip political bias

Nous Research retrained Alibaba's Qwen3.8-27B on 53 million tokens to produce Arete-27B, stripping out Chinese state propaganda. The model's propaganda rate dropped from 22 percent to under one percent while maintaining its general capabilities. The team found that the political bias is only a thin layer that can be removed through fine-tuning.

Original

State propaganda in some Chinese AI models is only "a thin layer" that fine-tuning can strip away

Nous Research retrained Alibaba's Qwen3.8-27B to produce Arete-27B, largely stripping out its pro-Chinese bias. The propaganda rate dropped from 22 percent to less than one percent, while general capabilities remained intact in tests. The method relied on 53 million tokens of supervised fine-tuning. The political bias is "applied late, it is applied thin," the team writes. The article State…

Read at the source
TechCrunch
Tools & products

Meta launches Spaces to organize tasks and chats with Muse agents

Meta has introduced Spaces, a feature enabling users to organize dedicated breakout conversations with its Muse AI agents around specific projects or topics. Users can describe an idea in the chat to create a specialized planner or tracker in the sidebar. The feature also supports integrations with third-party apps and data.

Original

Meta introduces Spaces, a new way to collaborate with your Muse agent

Meta's Muse agents continue to get new features and integrations. Spaces promises to be a new way to interact with the AI helpers.

Read at the source
The Verge
Policy & safety

China Reportedly Recruited Tech Executive to Spy on Nvidia CEO Jensen Huang

Chinese intelligence enlisted Megaspeed executive Alice Huang to spy on Nvidia CEO Jensen Huang. The Ministry of State Security sought to obtain information regarding American AI chips through the operation. Officials also aimed to potentially influence Nvidia to make its most advanced processors available to China.

Original

China reportedly recruited AI exec to spy on Jensen Huang

China's intelligence agency recruited Megaspeed executive Alice Huang to spy on Nvidia CEO Jensen Huang, according to The New York Times. The Ministry of State Security (MSS) enlisted her to glean information about American AI chips and "potentially influence the company to push for its most advanced chips to be made available to China." Megaspeed […]

Read at the source
Hacker News
Business

Microsoft Research Reorganizes Its AI Frontiers Group into Microsoft AI

The team developed tools such as the AutoGen framework and the Fara computer agent while sharing code across the organization. In early September, the group was reorganized and integrated into the broader Microsoft AI division.

Original

The end of Microsoft Research's AI Frontiers group

https://archive.is/AaY5O https://businessexaminer.ca/victoria-articles/item/microsoft... Comments URL: https://news.ycombinator.com/item?id=50406563 Points: 205 # Comments: 106

Read at the source
The Decoder
Research

AI agents fail benchmark for open scientific research questions

GPT-6 Astra achieved the best results among the tested models, but still fell short of the required threshold.

Original

Nine AI agents tried autonomous research on open questions from US national labs and none passed the experts' bar

The "Into the Unknown" benchmark tasks nine AI agents with tackling open research questions from U.S. national laboratories. None of the 135 submissions meets all requirements. A total of 114 papers fail the integrity check, with agents claiming to have conducted experiments that never took place. GPT-6 Astra performs best but still doesn't meet the experts' standards. The article Nine AI agents…

Read at the source
TechCrunch
Business

Google acquires unfinished Wyoming data center project from Crusoe

Crusoe has sold an incomplete data center project in Wyoming to Google, which plans to invest $17 billion to develop the site. Google assumed control of the facility after previously raising concerns about project costs and construction timelines. The transaction highlights the growing hurdles independent developers encounter when securing power and equipment for large-scale AI infrastructure.

Original

Crusoe sells troubled Wyoming data center project to Google

The deal highlights the challenges of developing massive AI data centers.

Read at the source
The Verge
Policy & safety

Concerns Raised Over US Approval of Nvidia AI Chip Sales to Chinese Firms

Concerns have emerged that the US Commerce Department and Nvidia are not properly vetting Chinese firms cleared to buy H200 AI chips. One approved firm, New Matrix, was found to have completely empty offices. Additionally, authorities have approved sales to Chinese entities with ties to blacklisted companies.

Original

The US let Nvidia sell chips to an empty office in China

Concerns are being raised that neither the Commerce Department nor Nvidia are doing their due diligence when it comes to vetting Chinese firms cleared to handle the company's H200 AI chips. According to The New York Times, New Matrix is on the list of approved firms, yet its offices are empty, "a hollow gray shell […]

Read at the source

Open-source repositories 8 repositories

Hugging Face model♥ 19,348
Models

Qwen/Qwen3.8-27B

It also features flexible thinking controls, allowing users to tune reasoning depth and retain reasoning context across interactions.

  • apache-2.0
  • #multimodal
  • #vision
  • #agents
Open model
Trending on GitHub★ 16,085
Coding

ParadevAI/Paradev

Paradev is a desktop application that allows developers to run multiple CLI coding agents in parallel using separate Git worktrees and branches. It integrates terminal sessions, code review, and browser previews into a single workspace powered by your own agent subscriptions.

  • typescript
  • MIT
  • #coding
  • #agents
  • #cli
Open repository
Trending on GitHub★ 14,709
Design & images

meta/astryx

It provides customizable theming without styling lock-in and allows teams to inspect or eject component source code directly into their projects.

  • typescript
  • MIT
  • #design-system
  • #react
  • #ui
Open repository
New on GitHub★ 7,636
Agents

wangzan101/Mika

Mika is an on-device AI agent for Android that operates apps and browses the web to complete tasks for you. It runs on a hidden background screen so you can keep using your phone normally, keeping your data local without relying on external project servers.

  • Kotlin
  • Apache-2.0
  • #agents
  • #android
  • #automation
Open repository about +794 stars a day
Hugging Face model♥ 5,392
Models

deepseek-ai/DeepSeek-V4.1-Flash-1017

DeepSeek-V4.1-Flash-1017 is an updated multimodal model that enhances performance across reasoning, coding, and visual benchmarks.

  • mit
  • #multimodal
  • #models
  • #vision
Open model
Trending on GitHub★ 4,043
Data & search

pipeshub-ai/pipeshub-ai

PipesHub is an open-source enterprise search platform and context layer that connects business tools like Slack, Google Drive, and GitHub. It enables both teams and AI agents to query company knowledge with citation-backed answers that respect existing user permissions.

  • python
  • Apache-2.0
  • #rag
  • #search
  • #mcp
Open repository
New on GitHub★ 2,939
Cybersecurity

qybaihe/bla

BLA is a defensive security tool designed to protect software builds, binaries, and app bundles from AI-powered reverse-engineering agents. It works by injecting decoy data and exhausting the analyzer's token and resource budgets to induce hallucinations without altering the application's actual runtime behavior.

  • TypeScript
  • MIT
  • #security
  • #agents
  • #defense
Open repository about +210 stars a day
Paper with code★ 5
Agents

THU-Agent/Coca-cola

This project compresses retrieval and tool outputs into latent representations instead of feeding long, raw text into LLM contexts. It provides code for training, evaluation, and baselines like LLMSummary and LLMLingua to reduce context overhead in multi-turn agent interactions.

  • #agents
  • #rag
  • #search
Open repository

Summaries are written automatically from the sources' headlines and excerpts and may contain mistakes. The full story is always at the source. To have a link or source removed, write to us from the contact section.

Let's start

Have an idea or a system that needs transforming?

Write us a few lines. Your message reaches us instantly, and we usually reply within one business day.

scouthub@lab: ~/contact

scouthub@lab:~$ contact --new --lang=en