The United States and China are locked in a race to develop increasingly powerful artificial intelligence models, a competition that takes on new significance as global discussions intensify around the risks posed by AI. While the U.S. has historically led in AI capabilities, research, and publications, China is rapidly emerging as an AI superpower, significantly closing the gap.
Experts and recent reports, including one from the Stanford Institute for Human-Centered Artificial Intelligence, indicate that China is challenging the U.S. lead in AI model development. Samm Sacks, a senior fellow at the Johns Hopkins School of Advanced International Studies, noted that the difference between U.S. and Chinese AI models is currently narrow and potentially fragile.
This technological race occurs against a backdrop of increasing AI-related concerns voiced by both nations. China's Minister of State Security, Chen Yixin, has warned that AI poses a range of threats, including risks to political and ideological security, cyberattacks, data leaks, and challenges to governance. He specifically cited the potential for AI-generated deepfakes to spread political misinformation and for AI to be used to identify and exploit cybersecurity vulnerabilities.
Despite these concerns, top Chinese leaders have called for accelerated AI development and technological self-reliance, while also emphasizing the need to control risks and establish a global framework for AI governance.
In the U.S., federal agencies have accused Chinese AI developers of engaging in "industrial-scale distillation," a practice where AI models are trained to copy or imitate the capabilities of more advanced U.S. models. The FBI, NSA, and CISA stated in September that Chinese firms had used unauthorized access and other methods to extract restricted capabilities from leading American AI models. U.S. officials have characterized these activities as "aggressive, malicious, and targeted."
China has refuted these accusations, asserting that distillation is a common technique used across the AI industry. Beijing has accused the U.S. of pursuing a monopoly in the AI sector and urged the U.S. to refrain from making unfounded claims.
Lizzi C. Lee, a fellow at the Asia Society Policy Institute’s Center for China Analysis, explained that while distillation is a standard industry practice, the U.S. agencies' specific allegations point to unauthorized access and methods used by Chinese firms to extract capabilities from sensitive American frontier models. This raises questions for the U.S. about how to control the outputs of AI, in addition to its efforts to control inputs like advanced chips.
Recent developments highlight China's rapid progress. Beijing-based startup Moonshot AI released Kimi K3, an AI model that surprised Silicon Valley with capabilities competitive with U.S. rivals. Although Kimi K3's ranking has fluctuated, it demonstrates China's ability to produce advanced models.
AI startup DeepSeek gained attention in early 2025 with its R1 model, which was noted for its cost-effectiveness compared to U.S. alternatives. In April, DeepSeek previewed its V4 model, boasting significant improvements in knowledge, reasoning, and "agentic" capabilities—the ability to plan and execute complex tasks with minimal human guidance.
Additionally, Z.ai's GLM-5.2, released in June, drew attention for its performance comparable to leading models from Anthropic and OpenAI. The subsequent release of GLM-5.3 in August further enhanced its coding and agentic abilities.
Alibaba also previewed its Qwen3.8-Max model in July, positioning it as its most powerful offering to date and a contender against leading U.S. AI systems.