6 Months to Live for Open Models

Nathan Lambert's July 12 essay separates two debates that could shape access to downloadable AI models: whether companies improperly copy competitors' capabilities, and how governments should handle models with dangerous capabilities. Lambert argues that restrictions proposed for the first problem could become a route to broadly limiting open models, affecting businesses that depend on running and adapting them. The essay is a policy argument: it explicitly acknowledges having no official information about the discussed executive order. Its six-month warning is Lambert's forecast, not an announced deadline. The distinction matters because copying disputes and capability risks require different evidence and policy choices.