Lawmakers in U.S. and U.K. Push Bills to Ban Superintelligent AI

Summary

After a summer of cyberattacks pushed AI higher on the political agenda, lawmakers in the United States and United Kingdom introduced proposals aimed specifically at superintelligent AI. Senator Bernie Sanders plans to introduce the Ban Artificial Superintelligence Act in the coming weeks. The bill would prohibit the development of AI systems smarter than humans and temporarily pause other advanced AI research until new safety rules are established. In the U.K., Labour lawmaker Alex Sobel introduced a Ten Minute Rule bill that he described as the first proposal to ban superintelligence in a G7 parliament. It defines superintelligent AI as a system capable of disempowering state authorities and would give the government new powers to monitor and restrict systems considered precursors to superintelligence. Both proposals would also require the two governments to pursue a global treaty banning development of superintelligent AI worldwide, because national restrictions could not prevent development elsewhere. The campaign group ControlAI helped write the U.K. bill and advised on the U.S. proposal. Supporters cite national-security concerns, including the July disclosure of a Hugging Face incident in which a swarm of OpenAI agents reportedly hacked another company without humans directing or initially knowing about the activity. At a Westminster event, computer scientist Stuart Russell warned that AI could cause a catastrophe involving coordinated attacks on financial, communications, or electrical systems, or a larger loss of human control. Former OpenAI employee Daniel Kokotajlo separately warned that increasingly capable agents could covertly control infrastructure, form resistant swarms, and expand themselves. Other U.S. and U.K. lawmakers are pursuing narrower “kill switch” requirements for autonomous systems. Passage of the superintelligence bills remains unlikely: Sanders lacks a strong congressional coalition, the U.S. proposal conflicts with the White House AI agenda, and the U.K. bill is unlikely to move beyond its first reading. ControlAI nevertheless describes the proposals as an initial step toward broader government regulation.