Chapter Twelve
When the Company Tried to Hold the Line
In July 2025, the U.S. Department of Defense awarded contracts of up to $200 million each to Anthropic, Google, OpenAI, and xAI for AI capabilities to address “critical national security challenges.” Anthropic’s contract included an acceptable use policy prohibiting use of Claude for mass domestic surveillance of U.S. citizens and for fully autonomous weapons systems — systems that, once activated, could select and engage targets without human involvement.
In January 2026, that became a problem. The DoD demanded unrestricted use of the technology for any lawful military purpose. Anthropic refused. Defense Secretary Pete Hegseth issued an ultimatum.
On February 27, 2026, President Trump directed all federal agencies to immediately cease use of Anthropic’s technology. Hegseth directed the DoD to designate Anthropic a supply-chain risk to national security and ordered all defense contractors — including Boeing and Lockheed Martin — to stop using Claude. Anthropic filed suit in the Northern District of California on March 9. On March 26, a federal judge issued a preliminary injunction temporarily blocking the designation and halted the President’s directive. The GSA restored Anthropic’s government contracts. The case continues in active litigation.
This is the most instructive episode in this book, and not because Anthropic was right or wrong. It is instructive because of what it shows about the architecture of the constraint.
A private company’s acceptable use policy is not a law. It is a contract term. Contract terms are enforced by courts. Courts are slower than operational deployments. The preliminary injunction took six weeks to obtain. In six weeks, a military contractor can train a workforce on a system, integrate it into operational pipelines, and create dependencies that a legal ruling can restore in principle but cannot undo in practice.
The state of your privacy and the limits on AI in autonomous weapons are being decided by contract negotiations and federal litigation between a technology company and the Pentagon. Not by Congress. Not by voters. By lawyers in a federal courthouse in California.
Anthropic tried to hold the line. It is holding it, so far, by a judge’s order. The question the case cannot answer is what happens when the deployment has already occurred before the injunction can be obtained — when the data has already been collected before the privacy ruling is issued, when the weapons have already acted before the autonomous weapons ban is enforced.
Two months after the judge’s preliminary injunction, Anthropic’s co-founder stood alongside Pope Leo XIV at the Vatican to present Magnifica Humanitas — the papal encyclical calling for AI to be ‘disarmed’ and warning that humanity faces a choice between constructing a new Tower of Babel and building a city where God and humanity dwell together. The same company that went to federal court that spring over autonomous weapons helped present the document calling for those weapons to be governed toward the common good. The litigation continues. The encyclical is in the appendix.