China requires MOFCOM approval to open-source models above 2T params or 10^26 FLOPS
zephyr_z9 · x · 2026-10-16
A newly circulated entry (266504X) in China's dual-use export control list covers frontier LLM technologies: models trained on important data, or whose core tech (weights, checkpoints, source code, architecture, training/inference optimization) involves ≥10^26 FLOPs, ≥200B parameters, or globally top-10 benchmark capability. Per the post, Chinese players wanting to open-source models above 2T parameters or trained on over 10^26 FLOPs will need MOFCOM permission. A carve-out exists for models trained on only trace (≤0.1%) important data with desensitization measures.
More from Safety
- Vitalik builds "paranoid mode" agent wrapper with Tor, zkAPI and local prompt rewriting — DavideCrapis · 2026-10-16
- Security team demos Signal 1-click exploit technique at Hexacon Paris — moyix · 2026-10-16
- China moves to impose registration and licensing on frontier AI model weights and chip designs — deanwball · 2026-10-16
- Binding agent approvals to exact content: a synthetic test for stale permissions — Crescitaly · 2026-10-16
- Red Teaming Production AI Agents: What Breaks Under Targeted Attacks — njyx · 2026-10-16
- Comprehension Audits: New Paper Proposes a Gate for AI-Written Code at Frontier Labs — ronbodkin · 2026-10-16