America's military AI ambition is colliding with its own feud and China's advance

A strategic capability is trapped inside a governance dispute

The military wants access to the strongest systems on a timetable measured in weeks, not ordinary procurement cycles. Frontier models, however, are built and updated by private companies that retain technical control, safety policies, and commercial leverage.

The Anthropic dispute turned that dependency into public confrontation. A supplier's red lines and an agency's demand for lawful-use freedom collided before Congress had established rules that could bind both.

China's coordination advantage carries a governance cost

A state able to align laboratories, infrastructure, data, and military priorities can move without the same public contractual fight. That can accelerate adoption and focus investment.

It can also erase independent refusal, civil-liberties safeguards, and public accountability. The democratic answer is not to copy command-and-control governance, but to make lawful decision-making faster and more credible.

Build speed on rules that survive a change of power

Congress should define prohibited and high-risk military AI uses, the required role of human judgment, audit access, incident reporting, and standards for domestic data. Procurement should reduce dependence on any single provider while preserving interoperability and test evidence.

Strategic resilience comes from institutions that can move quickly without inventing the rules during each crisis. A model lead can vanish in months; legitimate authority and reliable integration have to last longer.

• Put military AI red lines and reporting duties in statute.
• Require independent evaluation before classified deployment.
• Avoid single-vendor dependency through common interfaces and portable controls.
• Protect documented human refusal for consequential targeting and surveillance decisions.