AI models trained using 10^25 or more operations would be subject to an immediate development pause under new legislation proposed in Congress. The Ban Artificial Superintelligence Act, introduced by Senator Bernie Sanders and Representative Greg Casar, would permanently prohibit the creation of AI that exceeds human cognitive performance across most domains or has the capability to disempower humanity. The bill establishes a cabinet-level Department of Artificial Intelligence to create safety rules and mandate federal charters for developers, including government access to internal systems and staff. Violators could face a "corporate death penalty" involving asset and intellectual property forfeiture, and individuals may be sentenced to up to 20 years in prison.

The act forbids "superintelligence precursor" capabilities, such as the ability for models to automate AI research, resist shutdown, or evade human oversight. It grants the new federal department power to force flagged models offline and mandates that confirmed superintelligent systems be destroyed immediately. To prevent these technologies from emerging globally, the bill designates the pursuit of international agreements and AI-compute export controls as official US policy. Some critics argue that the legislation would destroy competition and primarily benefit dominant companies like OpenAI and Anthropic.