LiteLLM dominated security headlines (Check Point, TechRadar, Cybernews). NVIDIA doubled down on UK startup investment and European city agents. Science and open weights: Allen AI HiRO-ACEv2 on Hugging Face; China open-model gap narrows to ~7 months in one survey read.
Today
- When LLM gateways go wrong: LiteLLM findings — Check Point Research on exposed gateways and takeover paths. Check Point Research
- LiteLLM gateway flaws: millions of users at risk — TechRadar roundup of open-source AI gateway vulns. TechRadar
- LiteLLM vulnerabilities expose AI gateways — Cybernews on takeover and key theft. Cybernews
- HiRO-ACEv2: extreme weather at 3 km resolution — Allen AI sim model on Hugging Face. Hugging Face Blog
- Meta Muse Wonder vs OpenAI and Anthropic — TechGig on Meta’s agent push. TechGig
- China open AI models ~7 months behind US frontier — Startup Fortune survey read. Startup Fortune
- Intel inference microservices for Xeon + vLLM — KuCoin syndication of Intel launch. KuCoin
- OneAdvanced + NVIDIA for UK agentic AI — Business Wire on safer agents in critical sectors. Business Wire
- Mirascope: type-safe Python agent loops — Quasa on open-source LLM tooling. Quasa.io
- Top tools for serving open models on Kubernetes — HackerNoon ops roundup. HackerNoon
From NVIDIA
- New investments in the UK startup ecosystem — NVIDIA blog on deployment scale in Britain. NVIDIA Blog
- European cities deploy NVIDIA-powered AI agents — Urban operations and energy use cases. NVIDIA Blog
Watch: Open-weight
- Mistral “Le Chonk” trillion-parameter headline — Treat as unverified; Startup Fortune syndication. Startup Fortune
- AMD Ryzen AI Halo 192GB vs DGX Spark — Local RAM vs NVIDIA Spark speed tradeoff. Tech Times
- Jailbroken PS5 running a 27B model — Startup Fortune on edge hardware stunt. Startup Fortune
Built with utilities/ai_report_build.py (24-hour lookback). Edited for publication. About the AI Report.