China is shaping the future of open-source technology – including AI

Chinese AI companies such as Minimax made a big splash during the World AI Conference in Shanghai in July 2026. AP Photo/Ng Han Guan

China’s long road to open-source

The open-source movement has taken various forms, from collaborations aimed at bypassing corporate monopolies to business strategies for developing products quickly. Open-source software has allowed developers around the world to reuse and improve on shared code.

Linux, the core software powering Android phones and most of the internet, is an open-source technology. RISC-V, an emerging alternative to Intel and Arm chip designs, and next-generation mobile network interfaces are also open-source.

Early Chinese engagements with open-source aimed to develop technologies that were free from foreign control. A government-run software institute developed Red Flag Linux as an alternative to Microsoft’s Windows operating system, which the government saw as costly to maintain and vulnerable to foreign intelligence gathering. A Chinese military university later created Kylin, an operating system for the People’s Liberation Army and other government organizations.

AI could be a source of tension between U.S. President Donald Trump and Chinese President Xi Jinping when they meet in September. China Pool via Getty Images

In recent years, Chinese tech companies have embraced open-source technology as a way around sanctions and export controls. After the U.S. government blacklisted the telecom giant Huawei for its ties to the Chinese military, Google revoked the company’s access to licensed services and apps. But Huawei used Android’s open-source code to maintain its smartphone operating system. Eventually, Huawei developed an alternative called HarmonyOS NEXT and released its own open-source version as OpenHarmony.

Today, Chinese leader Xi Jinping is stressing the importance of science and technology independence and self-strengthening, known as keji zili ziqiang. The Ministry of Industry and Information Technology and other government bodies have supported domestic platforms, such as the state-guided OpenAtom Foundation, for open-source technology, and Gitee, a Chinese alternative to GitHub, a popular repository for open-source software. Domestic open-source operating systems such as openKylin and openEuler play a key role in securing China’s information technology infrastructure.

China is also positioning itself as an indispensable player in setting global standards for open-source hardware. For example, its work on RISC-V allows Chinese technology companies to reduce dependence on foreign intellectual property in semiconductors. Open hardware standards make Chinese companies less vulnerable to export controls and bans.

China’s strategy of “overtaking on a turn,” or closely following and then passing the front-runner, appears to be paying off. Chinese telecommunications companies, already leaders in 5G mobile network technology, are working with an international alliance to shape 6G, the next-generation standard for mobile communications. The Open-RAN standard the alliance is developing has complex implications for cybersecurity and strategic competition.

Open-source AI: With Chinese characteristics?

The contest over AI has moved beyond export controls of AI chips toward who leads open-source AI development. Xi is expected to discuss AI with U.S. President Donald Trump on his visit to the U.S. in September.

Major Chinese AI models such as Qwen, DeepSeek, Kimi and GLM are generally open-weight, not open-source, meaning that the model software and its final parameters are made available, but not the training data and methodology. AI companies often offer models in proprietary, open-weight and smaller open-source versions.

The White House and Silicon Valley remain split over the prospects of open-source AI. Anthropic and OpenAI have accused Chinese open-weight model makers of large-scale, unauthorized distillation, a technique used to copy proprietary AI models. The Trump administration and U.S. lawmakers have proposed restricting access to Chinese open-weight models, and some federal agencies and state governments have already moved to prohibit employees and contractors from using DeepSeek, for security reasons.

Users of Chinese AI face risks that are not widely understood. Chinese open-weight models are designed to align with government censorship and propaganda. While Chinese models are often distributed under standard licenses, such as MIT and Apache 2.0, some impose additional compliance obligations.

Moonshot’s Kimi K3 model license, for example, requires licensees exceeding specific revenue or user thresholds to “prominently” display the Kimi model name. MiniMax M3’s community license lists uses prohibited by “applicable laws or regulations,” raising questions over which jurisdictions apply.

Bilingual licenses developed under Chinese government guidance give Chinese users more access and Chinese courts more control. These include the Mulan Permissive Software License, version 2.0 and the OpenAtom Model License, version 1.0. Both are written in English and Chinese, but the Chinese version prevails in cases of conflicting interpretation, especially in Chinese jurisdictions.

Open-source technology has always been global

The U.S. regulates AI through a patchwork of federal and state laws. Internationally recognized frameworks may offer a better approach to governing open-source AI. Examples include the European Union’s Artificial Intelligence Act and multilateral initiatives such as UNESCO’s Recommendation on the Ethics of Artificial Intelligence, the Hiroshima AI Process and the International Standards Organization’s ISO/IEC 42001.

Open-source technology could also benefit from such governance.