Mania Africa 💎🌍

マイクロソフトのAIセキュリティ対策：AIのさらなる活用と略語のさらなる増加

Microsoft's solution to AI security: more AI and more acronyms #Register (Jul 27)

www.theregister.com/security/202...

They point to MCP, built by two engineers in London, that ended up under US-based governance.

https://itsfoss.com/news/british-flavored-linux-foundation-proposal/

DeepLens Diagnosis Agent: Agentic Workflow Design Lets a Small Reasoning Model Compete with Frontier LLMs

Mahmood Bayeshi et al.

#arXiv #cs.AI

Why blocking AI models won't stop the cyber threats they create cyberscoop.com/why-blocking...

#Microsoft escalates the AI security race with ‘Project Perception’ and a new in-house model www.geekwire.com/2026/microso...

[AP] A FLEX IN CORPORATE AMERICA, AI ‘TOKENMAXXING’ FADES AS WORKPLACES LOOK TO CUT TECH SPENDING

futurism.com/artificial-i...

AI agents are nudging pregnant women to anti-abortion sources. A handful of US tech firms are shaping intimate medical decisions – with little transparency about answers www.opendemocracy.net/ai-chatbots-...

OpenAIの「Hugging Face」攻撃後、テクノロジー大手各社が手を組んでOpenAIのモデルを称賛

Tech giants link hands to praise open AI models after OpenAI - Hugging Face attack #Register (Jul 27)

www.theregister.com/ai-and-ml/20...

Cybersecurity jobs available right now: July 28, 2026

📖 Read more: www.helpnetsecurity.com/2026/07/28/c...

#cybersecurity #cybersecuritynews #cybersecurityjobs

PoC exploit released for critical AD CS domain-takeover flaw (CVE-2026-54121)

Security researchers who discovered and reported CVE-2026-54121 (aka “Certighost”), a critical privilege elevation vulnerability in Active Directory Certificate Services (AD CS), have release…
#hackernews #microsoft #news

it's so goofy that they're trying to get investors to hodl on gen AI by constantly putting out stories about how it doesn't do what you tell it to because it's so damn smart. "We invented a digital sociopath! Please keep investing!"

Dario Amodei published Anthropic’s formal position on open-weights models on Sunday, breaking a silence that had drawn accusations that the company wanted to ban open-weight AI to protect...

🔗 https://thenextweb.com/news/anthropic-dario-amodei-open-weights-position-not-ban

Rethinking security for the age of #AI, blogs.microsoft.com/blog/2026/07...

#Microsoft launches its first cybersecurity model, plus a new agentic cybersecurity system, techcrunch.com/2026/07/27/m...

Googleがサイバー犯罪グループの新たな分類法を独自に発表

Google goes it alone with a new cybercrime crew taxonomy #Register (Jul 27)

www.theregister.com/security/202...

🟢 Thailand’s Ministry of Finance Attacked Using Hermes AI Agent

🗨️ Researchers at Hunt.io and security researcher Bob Diachenko discovered an unsecured infrastructure belonging to a hacke…

#news

AIBIX lab, a local company specializing in Generative Engine Optimization and AI performance benchmarking, unveiled a new 100-point “AI Brand Index” on Tuesday that measures how often and how favorably brands appear in the answers of generative AI chatbots.
www.koreajoongangdaily.com/business/seo...

Anthropic CEO Dario Amodei published a position paper on open-weight AI models. This comes after Nvidia issues its own statement in support of them. -

Cybersecurity Research, Development, and Demonstration for Energy Systems, www.energy.gov/ceser/cybers...

Microsoft launched MAI-Cyber-1-Flash inside MDASH, claiming 95.95% CyberGym performance with lower configuration cost and limited Azure AI Foundry private preview access.

Ranjan Veerabhadraswamy, Ajith Jubilson Emerson: Spectral Flow Certificates for Depth-Aware Long-Range Propagation in Graph Neural Networks https://arxiv.org/abs/2607.21607 https://arxiv.org/pdf/2607.21607 https://arxiv.org/html/2607.21607

Negi, Jain, Chakarborty, Negi, Shukla: ReCon: A Resource-Constrained Benchmark for LLM-Based Cybersecurity Compliance Across Ingestion and Retrieval Pipelines https://arxiv.org/abs/2607.22885 https://arxiv.org/pdf/2607.22885 https://arxiv.org/html/2607.22885

Vatsal Baherwani, Tom Goldstein, Ashwinee Panda: Not All LLM Reasoning is Visible in the Chain-of-Thought https://arxiv.org/abs/2607.22925 https://arxiv.org/pdf/2607.22925 https://arxiv.org/html/2607.22925

Siqi You, Bingsong Xu, Zhixian Zheng, Xinjian Peng, Yang Xie, Ying Wang, Jiarong Xu: Traceable LLM Reasoning for Fake-Order Fraud Detection https://arxiv.org/abs/2607.23075 https://arxiv.org/pdf/2607.23075 https://arxiv.org/html/2607.23075

Dipshikha Das, Arnab Banik, Md. Shariful Islam, Md Rayhanur Rahman: A Structured Cyber Threat Intelligence Dataset Using STIX 2.1 Entities and MITRE ATT&CK Mappings https://arxiv.org/abs/2607.23312 https://arxiv.org/pdf/2607.23312 https://arxiv.org/html/2607.23312

Hai Hu, Siyuan Song, Chongtian Shao, Kejia Zhang, Tianjian Zhu, Xiaojing Zhao: Reasoning or Memorization: Can LLMs Understand and Generate Chinese Xiehouyu Riddles? https://arxiv.org/abs/2607.23440 https://arxiv.org/pdf/2607.23440 https://arxiv.org/html/2607.23440

Ponte, Gibert, Kozak, Trizna, Pintor, Biggio, Roli, Demetrio: EXE-Bench: Ranking the Tradeoffs of AI-based Windows Malware Detectors for Real-World Usability https://arxiv.org/abs/2607.24177 https://arxiv.org/pdf/2607.24177 https://arxiv.org/html/2607.24177

Sofia Della Penna, Lorenzo Parracino, Luciano Pianese, Vittorio Orbinato, Roberto Natella: VulnGym: Evaluating Vulnerability Management Strategies against Advanced Persistent Threats https://arxiv.org/abs/2607.24552 https://arxiv.org/pdf/2607.24552 https://arxiv.org/html/2607.24552

Shuoyang Jasper Zheng, Anna Xamb\'o Sed\'o, Nick Bryan-Kinns: Explainable AI through the Lens of Material Agency: Enabling Musical Interface Design with Neural Audio Models https://arxiv.org/abs/2607.23309 https://arxiv.org/pdf/2607.23309 https://arxiv.org/html/2607.23309

Madan Baduwal, Priyanka Paudel: QFedPolyp: A Communication- and Inference-Efficient Federated Learning Framework for Polyp Segmentation https://arxiv.org/abs/2607.22743 https://arxiv.org/pdf/2607.22743 https://arxiv.org/html/2607.22743

Yu Xia, Zihan Lin, Wei Yang, Rui Zhong, Cheng Chen, Huan Ren, Yao Hu: LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation https://arxiv.org/abs/2607.24617 https://arxiv.org/pdf/2607.24617 https://arxiv.org/html/2607.24617

Kun Zhao, Xu Chen: Multi-Agent Privacy Game in Federated Learning: A Unified Mean-Field View https://arxiv.org/abs/2607.23029 https://arxiv.org/pdf/2607.23029 https://arxiv.org/html/2607.23029

Anthropic said the system is working as intended

futurism.com/artificial-i...

Youngseok Hwang, Joonsung Kwon, Geonwoo Lee, Hyunwoo Park: Domain-Prior-Regularized Graph Modeling for Anomaly Detection in Cyber-Physical Systems https://arxiv.org/abs/2607.23197 https://arxiv.org/pdf/2607.23197 https://arxiv.org/html/2607.23197

Bo Wu, Haoling Wang, Zhuodiao Kuang, Kateryna Shapovalenko: Blood Pressure Estimation from PPG: A Comparative Study of Direct and ECG-Mediated Deep Learning Pipelines https://arxiv.org/abs/2607.23406 https://arxiv.org/pdf/2607.23406 https://arxiv.org/html/2607.23406

L\'eo Hein, Giovanni De Nunzio, Aur\'elie Pirayre, Laurent Najman: Capacity-Aware Deep Learning for Generalizable Traffic Volume Estimation Across Links and Cities https://arxiv.org/abs/2607.24056 https://arxiv.org/pdf/2607.24056 https://arxiv.org/html/2607.24056

Jingwen Zhu, Keshu Wu, Pei Li, Steven T. Parker, Bin Ran, David A. Noyce: Forecasting the Emergence and Evolution of Crash Hotspots: A Unified Deep Learning Framework for Proactive Traffic Sa... https://arxiv.org/abs/2607.24168 https://arxiv.org/pdf/2607.24168 https://arxiv.org/html/2607.24168

Wenxuan Ma, Chaofan Zhang, Chao Xue, Yinghao Cai, Guocai Yao, Shaowei Cui, Shuo Wang: FeelWorld: Visuo-Tactile World Model for Hierarchical Contact Prediction and Planning https://arxiv.org/abs/2607.24267 https://arxiv.org/pdf/2607.24267 https://arxiv.org/html/2607.24267

Castan, Arora, Richard, Orsula, Pradalier, Olivares-M\'endez: Learning Adaptive Multi-Task Guidance, Navigation, and Control via Hypernetworks https://arxiv.org/abs/2607.24292 https://arxiv.org/pdf/2607.24292 https://arxiv.org/html/2607.24292

🤖 Certighost PoC published: authenticated attackers can hijack Windows domains via AD CS vulnerability. Exploit chain targets certificate services for full domain takeover.
https://www.bleepingcomputer.com/news/security/new-certighost-poc-exploit-lets-attackers-hijack-windows-domains/

Kaidi Wang, Daniel K C So, Zhiguo Ding: Age-of-Information Aware Federated Learning with Finite Speed Pinching Antenna https://arxiv.org/abs/2607.23595 https://arxiv.org/pdf/2607.23595 https://arxiv.org/html/2607.23595

We Europeans need to realise that China is not friendly or even neutral to Europe. Their conduct in Serbia and Hungary shows us that they hold our values, our respect of nature and our notions of worker and consumer rights in contempt.
China is an enemy of Europe.

autos.yahoo.com/policy-and-e...

Part 0 of my Math Behind Neural Networks series is now also on Medium, published in CodeToDeploy.

Same plain-language intro to what a neural network really is — now wherever you prefer to read.

medium.com/codetodeploy...

CVE-2026-16812 is a remotely exploitable VeloCloud Orchestrator OS injection flaw actively exploited in the wild, requiring only web access and no credentials.

155 CVEs patched?!

macOS版ClickFix：Appleユーザーが騙される手口

ClickFix on macOS: how Apple users are being scammed #Kaspersky (Jul 27)

www.kaspersky.com/blog/macos-c...

Since late March 2026, a global phishing campaign has used obfuscated JScript, Lua/AutoIt loaders, and fake .ttf files to deploy RATs and infostealers like Agent Tesla and Remcos on Windows. #LuaJIT #AutoIt #Windows

arXiv📈🤖
MOCA: A Transformer-based Modular Causal Inference Framework with One-way Cross-attention and Cutting Feedback
By Wang, Ghosh

Nvidia is reportedly in talks to provide financial backing for OpenAI's planned $250 billion data center project in Ohio. The backing is crucial as OpenAI has not yet achieved profitability. #AI #News

Microsoft unveils AI security tools it says outperform competing platforms
comments · posted on 2026.07.28 at 03:02:33 (c=0, p=3)

Anthropic opened its Economic Index to conversational queries via Claude. Aggregate, anonymised usage data — which tasks AI is taking, which occupations are adopting it, how it varies by sector. Same day as the $200M research fund. Worth querying before citing it as a source.

fortune.com/2026/07/27/a...

#SamAltman thinks the singularity is already here, but an expert says #OpenAI’s Hugging Face breach doesn’t prove it

fortune.com/2026/07/27/s...