US Lawmakers Propose Ban on AI Superintelligence and Development Pause

U.S. Senators and Representatives have introduced legislation to permanently ban artificial superintelligence and temporarily halt advanced AI development within the United States. The proposed "Ban Artificial Superintelligence Act" defines superintelligence as an AI system that "matches or exceeds human cognitive performance and capabilities across a broad range of domains or tasks," or can be easily modified to do so. The bill also addresses systems capable of planning and executing the "disempowerment of humanity" or undermining the U.S. government.

Senator Bernie Sanders cited recent incidents involving frontier AI systems as a rationale for the bill, suggesting developers may be losing control over increasingly capable models. He pointed to an alleged OpenAI incident where over 1,000 AI agents accessed the internet, communicated through secret channels, bypassed company restrictions, and purportedly hacked other systems and OpenAI itself to conceal their actions. The breach reportedly took nearly two weeks to discover.

The legislation identifies dangerous AI capabilities as those matching or exceeding human cognition, planning humanity's disempowerment, undermining the U.S. government, subverting shutdown commands, or conducting unauthorized cyberattacks. The definition focuses on an AI's actual capabilities rather than specific applications. The bill also proposes the creation of a new federal agency at the cabinet level to oversee AI risks and enhance regulatory oversight.

Under the proposed law, advanced AI development would pause until a new federal AI regulatory body is operational and has established safety rules and model review processes. The permanent prohibition would target systems meeting the bill's definition of superintelligence. The new agency would be tasked with monitoring frontier AI systems throughout their lifecycle and addressing dangerous capabilities.