When the Have-Nots Become the Haves
Techstrong.ai, Thursday, July 9th, 2026
China's shift from open-source AI to potential export restrictions mirrors how underdogs embrace openness then close it.
Alan Shimel examines how China's AI strategy follows a historical pattern where underdogs embrace openness, then restrict access after gaining leadership.
Chinese firms like DeepSeek, Qwen, and Zhipu released open models aggressively while trailing US rivals, but now that they compete, Beijing is weighing export restrictions on advanced models. Shimel calls this rational incentive alignment, noting frontier AI is strategic infrastructure like semiconductors, not ordinary software.
Unlike Linux or Terraform, frontier models cannot easily be forked because they require training infrastructure and data pipelines, making the open ecosystem fragile. Both the US and China now treat advanced AI as a geopolitical asset.