The bill, titled the "Human-Controlled AI Act," would prohibit models from recursively self-improving or autonomously modifying core objectives, containment, or shutdown controls until federal safeguards are in place and approved by an agency. A new federal agency would be responsible for developing safety regulations, a licensing system, audits, and safety standards. The bill also requires AI companies to hold extensive liability insurance and calls for government export controls to prevent adversaries like China from developing dangerous AI systems.