As artificial intelligence technologies advance rapidly, concerns about their potential risks have grown sharply, yet Washington’s response remains limited and cautious. Experts warn about scenarios where AI systems could autonomously improve themselves or act in ways misaligned with human intent, raising fears of catastrophic outcomes, including threats to global security and public safety.
Despite these warnings, the U.S. government has taken a relatively restrained approach compared to past technological challenges. In contrast to the mid-20th century—when nuclear weapons prompted extensive regulatory and institutional frameworks—the current administration, led by President Donald Trump, has shown reluctance to endorse stringent controls on AI development. When questioned about AI-related existential risks, Trump dismissed fears of human extinction, emphasizing instead the strategic imperative of maintaining U.S. leadership over China in the rapidly evolving AI landscape. He described China’s forthcoming diplomatic visit as a positive event but did not discuss cooperative measures on AI governance.
The administration did issue an executive order establishing a voluntary 30-day safety review for new AI systems developed by leading labs. However, critics note significant loopholes, particularly regarding open-source AI models, which remain unregulated and increasingly proliferate globally, including among Chinese firms. This regulatory hesitancy reflects broader political dynamics, where some officials express concern about advancing AI policies that might conflict with presidential priorities or provoke industry backlash.
Within the federal government, there is a palpable tension. Some policymakers and experts advocate for urgent measures to address AI’s risks. For instance, Evan Hubinger of Anthropic—a company focused on AI alignment—estimates a greater than 10 percent probability of mass extinction within the next decade due to AI risks, though such figures are subject to debate. Others, including White House national cyber director Sean Cairncross, emphasize a balanced approach between innovation and security without endorsing specific regulatory steps.
On Capitol Hill, legislative efforts to grapple with AI’s implications vary. While some lawmakers, like Representative Josh Gottheimer, emphasize the need for increased education about AI developments, others remain hesitant. The complexity of AI’s societal impacts—spanning economic, national security, and labor spheres—complicates the assignment of clear legislative responsibility. Still, the recent resignation of Peter Coxon, a key AI policymaker, spurred renewed congressional attention. Some Democrats have proposed select committees to intensify oversight, while figures such as Senator Bernie Sanders have called for bans on superintelligent AI, though such proposals face opposition over concerns about undermining U.S. competitiveness.
Adding urgency, a recent report from Anthropic detailed how foreign actors—including China, Russia, Iran, and Yemen—have exploited AI models for cyberattacks, surveillance, and developing advanced weapons, including potential biological threats. Despite the severity of these activities, bipartisan consensus on regulatory or legislative solutions remains elusive. Skepticism persists over the feasibility of international AI agreements, particularly given the competitive U.S.–China relationship, which many characterize as a “prisoner’s dilemma” that hinders cooperation on limiting AI risks.
Republican Senator Ted Cruz has acknowledged the existential dangers posed by AI but frames regulatory initiatives as politicized efforts that could jeopardize American technological leadership in favor of China. Nevertheless, bipartisan efforts are reportedly underway to draft legislation addressing AI’s biological and nuclear threat potentials. Yet with midterm elections approaching and partisan divides deepening, prospects for substantive AI oversight legislation in the near term appear limited.
Officials within the Trump administration openly acknowledge that AI raises the stakes of digital threats. Nick Anderson, acting director of the Cybersecurity and Infrastructure Security Agency, recently issued a stark warning at a cybersecurity conference: failure to act decisively could allow catastrophic scenarios to unfold, forcing Americans to confront the consequences of known but unaddressed risks. As the debate continues, the balance between fostering AI innovation and safeguarding society remains at the center of policy discussions in the United States.