The Department of War canceled a $200 million contract with Anthropic on Friday and officially labeled the AI company a supply chain risk. The breakdown occurred because Anthropic refused to remove terms of service restrictions regarding mass domestic surveillance and autonomous weapons. Within hours of the collapse, the Pentagon announced a new agreement with OpenAI to deploy its models on classified military networks.
We are witnessing a bizarre collision between national security and the modern software licensing model. For years, the tech industry has forced buyers into a paradigm where you no longer own the tools you purchase (you just rent access subject to the vendor dictating acceptable use). I respect the moral stance Anthropic is taking here, but I would much prefer a return to a world where a purchase means actual ownership. Microsoft does not call the Pentagon to audit how they use an Excel spreadsheet. If this helps moves things for consumers back towards the old model, that would be a huge win but I’m not holding my breath.
The government response to this way more troubling. Labeling a domestic AI lab a supply chain risk simply because it will not alter its terms of service screams of authoritarian overreach. If a vendor refuses to meet your operational requirements, the free market dictates that you simply take your budget to a competitor (which the Pentagon literally did the exact same day). Threatening to destroy an American company with national security designations over a standard contract dispute is a terrible and dark precedent.