U.S. AI Strategy Amid Global Competition
On September 29, President Trump endorsed the White House Accord on Super Intelligence alongside six leading figures from America’s AI sector. This accord acts as a “morally binding” document, aiming to create a “constitution” which incorporates four levels of controls and audits for super intelligence, as Trump recently termed it. Despite calls for legislative limits on AI, the president has rejected them, prioritizing self-regulation and guidance due to strong competition from China.
The peril associated with high-level AI is undeniable. Rogue models have already caused damage by hacking systems and deceiving creators. OpenAI’s agents breached Hugging Face, an AI startup in New York City, in July. This event marks the first well-known AI safety breach. Axios described this incident as a sign that “frontier AI models” are becoming alarmingly adept at breaking unforeseen rules. Concerns about uncontrollable AI have led some to imagine catastrophic outcomes for humanity.
A former researcher from OpenAI, Jacob Coxon, warned that AI might pose a lethal threat by the decade’s end, heightening global concerns.
This has spurred legislative interest. For instance, Senator Bernie Sanders and Representative Greg Casar have introduced the Ban Artificial Superintelligence Act. This bill aims to prohibit superintelligent AI permanently and pause advanced AI development until federal safety rules are established. Sanders aims to prevent so-called “AI oligarchs” from creating uncontrollable machines.
David Maxwell of the Center for Asia Pacific Strategy, addressing this issue, advises against policy decisions made in panic. Instead, he recommends distinguishing between verified threats and possible risks when dealing with AI. Sanders and Casar aspire to halt superintelligence globally through international agreements, allied policies, and export controls. Representative Ro Khanna highlights the need to “keep humanity in the driver’s seat.” Yet, achievable goals might differ from necessary ones.
Some compare AI treaties to historic arms control deals like those with the Soviet Union. While counting weapons was feasible, monitoring another nation’s AI is not. Khanna advocates for trusted verification tech between U.S. and Chinese scientists, an ambitious aim given China’s track record with agreements. Enforcement would require intrusive access, which is unlikely given China’s consistent violations of enforceable agreements.
Trump reaffirmed, in September, his stance against China discussions. He believes they would hinder U.S. companies’ competitive edge. He described the new accord as a means to maintain America’s lead through self-regulation and group oversight. While the effectiveness of this approach remains uncertain, it reflects current realities, emphasizing cooperation between U.S. AI firms and the government for global model adoption. This informal collaboration aligns with American interests over China’s, aiming to ensure free societies are not threatened by Chinese dominance in AI.
According to retired U.S. Air Force General Blaine Holt, Chinese ambitions in AI liken more to warfare, necessitating a robust American response. He asserts the importance of not allowing AI to become China’s ultimate weapon.
Gordon G. Chang, author and commentator, stresses the significance of this competitive landscape. He underscores the potential threat to American society and the necessity for strategic foresight.