AI News Today (8 Nov 2026): Chinese Intelligence Reportedly Recruited Exec to Spy on Nvidia CEO
Today in AI: Chinese Intelligence Reportedly Recruited Exec to Spy on Nvidia CEO; How to Run NVIDIA Dory 4B with llama.cpp; Questions Raised Over Nvidia H200 Chip Sales to Empty Chinese Office.
Today's briefing covers major developments in AI infrastructure, geopolitical scrutiny, and new open-source tooling. We examine everything from chip supply oversight to advanced inference control planes.
1. Chinese Intelligence Reportedly Recruited Exec to Spy on Nvidia CEO
An intelligence memo circulated earlier this year alleges that China's Ministry of State Security recruited a female tech executive to spy on Nvidia's CEO. The operation highlights ongoing international espionage concerns surrounding the artificial intelligence sector.
Why it matters: It underscores the high-stakes geopolitical tensions surrounding key executives and intellectual property in the artificial intelligence industry.
Read the full story at TechCrunch
2. How to Run NVIDIA Dory 4B with llama.cpp
A new coding guide details how to build the pinned llama.cpp branch for NVIDIA's Dory 1.0 4B looped language model. Developers can analyze compute costs, test memory caching, and execute paired evaluations.
Why it matters: Provides developers with actionable steps to deploy and benchmark advanced looped language models locally or in custom pipelines.
Read the full story at MarkTechPost
3. Questions Raised Over Nvidia H200 Chip Sales to Empty Chinese Office
Concerns have emerged regarding whether the Commerce Department and Nvidia performed adequate due diligence on Chinese firms cleared to receive H200 AI chips. Reports indicate that an approved firm named New Matrix currently operates out of an empty office.
Why it matters: Highlights potential regulatory and oversight gaps in compliance policies governing the export of high-performance hardware.
Read the full story at The Verge
4. Buoyant Releases Kinference Open Source Inference Control Plane
Buoyant has launched Kinference, an Apache 2.0 inference control plane designed for vLLM on Kubernetes. The tool introduces declarative routing policies, KV-cache-aware load balancing, and per-tenant metrics, which cut median time-to-first-token significantly in benchmarks.
Why it matters: Offers teams scaling LLMs on Kubernetes a robust way to optimize routing, caching, and infrastructure performance.
Sources
- Megaspeed exec was recruited to spy on Nvidia CEO, according to US intelligence memo · TechCrunch
- How to Run NVIDIA Dory 4B with llama.cpp: Loop by Loop Guide · MarkTechPost
- The US let Nvidia sell chips to an empty office in China · The Verge
- Buoyant Releases Kinference: An Open Source Inference Control Plane for Kubernetes and vLLM · MarkTechPost
This article was written with AI assistance from the sources above. Facts are attributed to their original publishers.
Written by Jayesh Patil
AI & full-stack developer. I build RAG assistants, AI automations and fast websites for businesses.