Anthropic Labeled Supply Chain Risk by Pentagon

Dwarkesh Patel published a sharp essay on the Department of War designating Anthropic a "supply chain risk" after the company refused to strip safeguards around mass surveillance and autonomous weapons. The core tension is real - 100 million CCTV cameras already exist in the US, and processing all that footage with current LLM pricing would cost around $30 billion a year, a number dropping by an order of magnitude annually. Patel argues that government coercion of frontier labs may be pointless anyway, since open-source models will eventually reach the same capability tier. Prediction markets give 81% odds the restriction gets reversed. The deeper question Patel raises - to whom should AI systems actually be aligned - is one that Anthropic, OpenAI, and Google have all carefully avoided answering directly.