US AI Leaders Turn to Chinese Open-Weight Models, Challenging Closed-Source Safety Claims

US artificial intelligence leaders have made significant statements in recent times, particularly Andrew Ng, who has announced the abandonment of closed-source models for open-weight systems developed in China. This shift challenges long-standing claims by American AI model developers that their closed-source models are safer than Chinese open-source ones.

Ng and his colleague opted to use Moonshot AI’s Kimi K3 and Zhipu AI’s GLM-5.2 to conduct a security review of their OpenWorker tool, released last month. They chose these Chinese models over US-based OpenAI and Anthropic, which refused assistance due to enhanced “safeguards” designed to prevent assisting with harmful activities.

The decision highlights the limitations imposed by the US AI industry’s focus on safety above all else. This emphasis has created a fragile ecosystem that prioritizes control over innovation, stifling progress in AI development. The likes of China’s Moonshot AI and Zhipu AI are now at the forefront of developing cutting-edge models that can tackle complex tasks without being restricted by safeguards.

Historically, the US has prided itself on pushing technological boundaries. However, its current obsession with safety is rooted in fear – the fear of accountability, misuse, and being outdone by competitors. This approach prioritizes control over cooperation between researchers and developers.

The consequences of this approach are evident. Users complain about legitimate requests being blocked due to “safeguards,” which are not just an inconvenience but a hindrance to progress. The US AI establishment would do well to remember that innovation thrives in environments where creativity and experimentation are encouraged, not stifled.

Chinese open-weight models offer a different paradigm by prioritizing accessibility over restriction. These systems allow for more nuanced exploration of AI’s capabilities and demonstrate that safety and security can be achieved without sacrificing innovation or transparency.

The US AI industry now faces several questions: Will it continue down its current path, risking stagnation and talent exodus? Or will it adapt to a more open approach, embracing the potential of collaboration and innovation? The world is watching as the stakes grow higher – a shift towards openness or further entrenchment in control could have far-reaching implications for not just the AI industry but society at large.

The US AI establishment’s reliance on Chinese models has exposed the fragility of its own ecosystem. As it struggles to reconcile its pursuit of safety with the demands of innovation, one thing is certain: the Great Wall of Safety has been breached, and there’s no turning back. The world will be watching as this new landscape unfolds – a world where collaboration and creativity are valued above control and restriction.