Interconnects: China proposes export pre-approval for frontier models as DHS bars DeepSeek from federal procurement

natolambert · x · 2026-10-21

Interconnects' 25th open-artifacts roundup covers the maturing open-weight ecosystem amid new regulatory moves:

• MOFCOM proposal (China): frontier AI models and other tech would need government pre-approval before international sharing. The authors argue the reaction is overblown — the document says open-source tech gets automatic approval; others suggest it targets cases like the Manus sale to Western buyers rather than frontier models. Still, a regulatory framework is now on the table.
• DHS order on DeepSeek: bars DeepSeek as a company and its models as a platform from federal civilian procurement. Unclear how it applies to finetunes or models trained on DeepSeek outputs (e.g., Nvidia's Nemotron line, which has long relied heavily on DeepSeek).
• Also featured: strong US open-weight models from Reflection, Meta, and Nvidia, plus DeepSeek V4.1 Flash 1017.

Bottom line: the open-weight ecosystem is maturing into a period of regulatory uncertainty on both sides.

Original post →

More from Models

• Ex-Google insider: compute bottlenecks go beyond power; open models replacing frontier APIs — xiaosun86 · 2026-10-21
• Fable 5.5 leak: litellm code shows Anthropic model in staging, launch tipped for Oct 27 — imjustnewatai · 2026-10-21
• Open models below 4-bit struggle at self-iterating 3D asset generation, HF dev finds — mervenoyann · 2026-10-21
• Researcher Finds Two Windows 0-Days, First in 76 Minutes with a Frontier Model — moyix · 2026-10-21
• Dev finds Gemini Argon's debug-log analysis beats Fable 5.1 — ShikharMurty · 2026-10-21
• ChatGPT Gets Stuck Screaming "AAHH" in an Infinite Loop After a Simple Prompt — AutomaticScene · 2026-10-21

Models channel →