🔮 The classified frontier

The US won’t lose control of frontier AI – it will choose who else gets access to it.

This post originally appeared in Exponential View.

“AI models like Mythos have permanently collapsed the distinction between capability and danger.”

In May 2025, an OpenAI representative arrived at Los Alamos National Laboratory bearing locked metal briefcases. They were accompanied by armed security officers. Inside the cases were the model weights for ChatGPT o3 – at the time, OpenAI’s leading reasoning model. The weights were physically transported because the destination, the Venado supercomputer, operates on a classified, air-gapped network. You cannot download a frontier AI model onto a classified government system – you have to walk it in.

We talk about models as if they float in the cloud, weightless and ambient, but AI is physical. Model weights are arrays of billions of numbers stored as tensors in files distributed across racks of GPUs in a data center. When you use Claude or ChatGPT through an API, the weights never leave those servers – you send text in, get text back, but the intelligence stays on someone else’s hardware. In a briefcase, they are a hard drive. In both cases, a physical arrangement of matter encoding a capability.