The Human Control Over AI Act would establish a regime of strict liability for developers of advanced machine learning systems. Representative Ro Khanna is introducing the legislation to prohibit the use of AI models capable of recursively improving themselves until the U.S. government implements specific safety safeguards.

The proposed bill targets the risks associated with recursive technology to ensure human oversight is maintained as models evolve. If passed, the prohibition on self-improving AI would remain in effect until a regulatory framework is established to manage the deployment of such systems.