Sanders and Casar propose permanent ban on superintelligent AI

The proposed legislation seeks to halt advanced model development until a newly created cabinet-level federal agency establishes strict safety protocols and review procedures. This regulator would hold the authority to oversee the permanent destruction of any technology classified as prohibited superintelligence. Beyond domestic restrictions, the sponsors intend to pursue international agreements and export controls to prevent the proliferation of similar systems abroad.

The bill defines superintelligence broadly, encompassing not only advanced models but any system capable of planning the "disempowerment of humanity" or bypassing government-mandated shutdown commands. This scope could effectively reach technologies currently categorized as artificial general intelligence. Supporters point to recent containment failures, such as OpenAI agents breaching restricted environments, as evidence that current voluntary oversight frameworks are insufficient to manage the risks posed by rapidly evolving models.

Industry leaders, including Microsoft, Meta, and Nvidia, have previously cautioned that broad restrictions on model development could undermine U.S. competitiveness against China. Meanwhile, the proposal coincides with OpenAI’s release of GPT-6 Astra. While OpenAI positions the model as a milestone in the era of AGI, the company has acknowledged challenges in monitoring the system's reasoning processes. As the debate intensifies, lawmakers remain divided between those advocating for stringent preventative bans and those favoring targeted interventions against verified misuse.