The Importance of US Leadership in Open-Source AI Development

Joseph F. Dunford, Francis F. Townsend, and Michael J. Morell

Source Link

The United States is in a race it cannot afford to lose. China is rapidly gaining ground in artificial intelligence (AI), utilizing open-source models to disseminate powerful, low-cost AI systems rooted in authoritarian values globally. These models aren’t just competing – they’re starting to win, and if left unchecked, they could become the foundation of the world’s digital infrastructure. To protect our national security, economic future, and democratic ideals, the United States must ensure that it leads in both open- and closed-source AI development.

China’s Open-Source AI Models Are Surpassing US Benchmarks

China’s AI gains aren’t just hypothetical. Its open-source models now rival or outperform top US systems. Alibaba’s Qwen 2.5 has beaten leading American models on key benchmarks, while DeepSeek’s open-source R1 made headlines earlier this year and is expected to be followed soon by an even more advanced R2 model. These breakthroughs are no accident: they’re part of China’s national strategy to become a global tech leader and spread censorship-driven AI across the globe. As the US Intelligence Community 2025 Threat Assessment warns, China aims to dominate AI by 2030, with its tech sector projected to drive nearly a quarter of its economy by 2026.

The US Must Strengthen Both Closed and Open-Source AI to Compete Globally

To counter the adoption of models like DeepSeek across the world, the United States must offer accessible alternatives that are powerful, low-cost, and – critically – underpinned by American values.

Many of the best-known US large language models – such as Anthropic’s Claude, OpenAI’s ChatGPT, and Google’s Gemini – were developed as closed models, meaning their underlying technology remains proprietary. The benefits of these models include control of how the technology is used and deployed. Closed models also protect confidential intellectual property (IP), algorithms, and data from theft or misuse, including from foreign adversaries.