The AI race America can actually win The United States should shift its artificial intelligence strategy from a race for centralized scale, which favors China, to a distributed architecture that empowers individuals. This approach plays to America’s strengths in open markets and democratic institutions, allowing AI to operate locally on trusted devices and networks. Such a decentralized system is crucial for military effectiveness in degraded or disconnected environments, offering resilience and operational flexibility that authoritarian competitors will find difficult to match.

• The US and China are discussing AI guardrails, balancing innovation and safety.
• A centralized AI approach, focused on scale (data centers, chips, models), plays to China’s strengths.
• China can align state power, industrial policy, data, energy, surveillance, and capital more effectively than a democracy.
• The US can win by adopting a distributed architecture where AI models are controlled by individuals.
• Distributed AI enables AI to be embedded in trusted devices and local networks, with the cloud as a support layer.
• This approach allows for adaptation, improvement, and use of trusted AI without central direction.
• The military is a prime example, requiring an edge-first, AI operating locally across units and platforms.
• Forces need AI capabilities that work in degraded, disconnected, and high-risk conditions, independent of constant cloud access.
• Advancements allow AI to run on smaller, cheaper, and more portable devices.
• Specialization is key: users need AI tailored to their specific missions, data, and context, not a single all-purpose model.
• A distributed network of locally controlled, specialized AI models offers reliable operational value when communications fail.
• China prioritizes centralized control; America’s advantage should be accountable empowerment through resilient, decentralized systems.