Democratic Congress prioritizes banning recursive self-improvement in AI

Representative Greg Kassar, chair of the Congressional Progressive Caucus, outlined a 90-day priority for the incoming Democratic majority to pass legislation banning recursive self-improvement (RSI) in AI. Kassar co-sponsored the Ban Artificial Superintelligence Act with Senator Bernie Sanders, which aims to prevent the development of AI more powerful than humans can control by restricting AI from deceiving humans or creating weapons of mass destruction. He proposes a federal cabinet-level department to monitor frontier labs like OpenAI, Anthropic, and Google DeepMind, measuring compute usage and preventing AI from developing other AI models. Kassar argues that current lobbying by AI executives, who cite the China threat to justify continued racing, is dangerous and that the US must end the AI arms race through verifiable international agreements rather than unilateral acceleration. He also proposed a tax on big AI companies to fund a jobs program, aiming to disincentivize job displacement and redirect compute resources toward medical research and climate solutions.