AI News Summary 2026-08-18

Also available in:

Español

.. lang: en

AI News Summary — August 18, 2026

GAFAM and Major AI Companies

  • OpenAI published The Defender’s Window, a security-focused article about how AI is reshaping cybersecurity for attackers and defenders alike. Source: OpenAI
  • OpenAI also published New policy ideas for the Intelligence Age, which states that the company is funding 14 independent projects exploring new AI policy ideas. Source: OpenAI

Influencers and Tech Blogs

  • Simon Willison highlighted Qwen 3.8 27B scores 52 on the Artificial Analysis Intelligence Index, noting that it matches GPT-5.6 Luna (max) and trails Qwen 3.8 235B by just one point. Source: Simon Willison
  • Simon also linked to a 404 Media investigation about a shipment of rare books that reportedly ended up at an Amazon AI training facility. Source: Simon Willison

Generative Imaging

  • Midjourney’s updates page currently displays a banner reading V8.2 is live! and links to the new version announcement. Source: Midjourney updates

Chatbots and Agents

  • Hermes Agent v0.20.4 was released on August 18, 2026, with a broad patch covering desktop transparency, the Sessions/Bots sidebar, Bot Mode fixes, skill-install advisory scanning, cron media-send hardening, and SessionDB fixes. Source: GitHub release

On-premises AI and serving

  • vLLM published Distributed Layerwise Offload: Scaling Toward 200B+ DiT Models Efficiently in vLLM-Omni, describing sharded/streamed DiT weights and a measured 124 GB Cosmos3 model on 64 GB HBM. Source: vLLM
  • The latest release of llama.cpp, b10472, is still on the local-serving watchlist and remains relevant for people running models locally. Source: GitHub release