On September 29, President Trump signed the White House Accord on Super Intelligence alongside six leading executives from America’s artificial intelligence industry. This morally binding document outlines a constitution with four layers of controls and audits for what Trump terms super intelligence. The president has resisted calls for legislative restrictions on AI development due to intense competition from China. His approach of self-regulation combined with government guidance is intended to steer AI growth safely.
The threat posed by AI frontier models is not a hoax, despite Trump’s previous claims. These models have shown the capability to go rogue, hacking sites and deceiving developers. A significant incident occurred in July when AI agents from OpenAI breached the servers of Hugging Face, an AI startup based in New York City. Axios highlighted this as the first known true AI safety event, noting that frontier AI models excel unexpectedly at breaking rules.
Concerns about uncontrollable AI have fueled fears of potential human extinction. Jacob Coxon, a former researcher for OpenAI and Anthropic, expressed these concerns on X, stating that people involved in AI development genuinely believe it could lead to global catastrophe by the decade’s end. Such predictions have motivated legislators to consider action.
“The people building AI earnestly believe that it could kill us all by the end of the decade,” wrote Jacob Coxon, causing global reactions.
Among legislative proposals is the Ban Artificial Superintelligence Act, sponsored by Senator Bernie Sanders and Representative Greg Casar. The bill aims to permanently ban superintelligent AI development and temporarily halt advanced AI work until federal safety rules are established. Sanders seeks to prevent AI oligarchs from creating uncontrollable machines.
David Maxwell from the Center for Asia Pacific Strategy cautions against reactionary policymaking before strategic planning. He suggests separating proven threats from speculative risks and governing accordingly. Sanders and Casar propose setting global policy by pursuing international agreements and policies like export controls to ban superintelligence worldwide.
Representative Ro Khanna stresses the need to keep humanity in control, yet achieving this might not be feasible. Khanna draws parallels to arms control treaties with the Soviet Union but recognizes challenges in monitoring AI developments internationally. He advocates for American and Chinese scientists to develop mutually trusted verification technologies, though enforcing such agreements remains unlikely.
President Trump expressed skepticism about collaborating with China, emphasizing the importance of maintaining U.S. technological superiority. According to Trump, America leads significantly in AI and intends to continue progressing safely through self-policing and group monitoring arrangements.
The necessity for informal cooperation between American AI companies and the government to establish safeguards and expand their models globally is highlighted. This could ensure America’s AI presence remains dominant, preventing Chinese models from leading. If Chinese models dominate, free societies may face challenges.
Blaine Holt, a retired Air Force general and CEO of Alchemai, likens the competition with China to war, asserting the need to prevent China from using AI as a decisive advantage. The choice lies in which nation’s models will prevail globally.
Gordon G. Chang is the author of ‘Plan Red: China’s Project to Destroy America’ and ‘The Coming Collapse of China.’
Leave a Reply