Defense contracts expose model surveillance loopholes

When large AI firms sign agreements with defense or intelligence agencies, contract wording can create surveillance, control, or data‑access loopholes that quickly become public controversies. Independent technical audits and community analysis (e.g., on LessWrong) are emerging as the main mechanism to find and pressure‑fix those gaps. — This matters because private–public AI procurement is creating new governance fault lines where corporate policies, national security interests, and public accountability collide.