Donald Trump and Elon Musk’s tumultuous strategies for reform are disrupting governmental operations. Essential activities have been stopped, tens of thousands of federal employees are being urged to leave, and congressional directives are being ignored. The subsequent phase: The Department of Government Efficiency allegedly aims to utilize AI to reduce expenses. Per The Washington Post, Musk’s team has commenced analyzing sensitive information from governmental systems using AI programs to assess expenditures and identify potential cuts. This may result in the removal of human positions in favor of automation. As one government official monitoring Musk’s DOGE team remarked to the Post, the overarching objective is to deploy AI to substitute “the human labor force with machines.” (Representatives for the White House and DOGE did not reply to inquiries for comments.)
Enhancing government efficiency through AI is a commendable endeavor, and this concept is not novel. The Biden administration revealed over 2,000 AI applications in progress throughout the federal sector. For instance, FEMA has begun employing AI to assist with damage evaluations in disaster-stricken areas. The Centers for Medicare and Medicaid Services have initiated the utilization of AI to detect fraudulent billing. However, the notion of supplanting dedicated and principled civil servants with AI entities is both novel and intricate.
The civil service—the extensive group of individuals who manage government agencies—plays an essential role in implementing laws and policies within society. New presidents can issue significant executive orders, yet they often lack real impact until they induce changes in the actions of public officials. Regardless of whether you view these individuals as crucial and inspiring altruists, monotonous bureaucratic operatives, or as representatives of a “deep state,” their substantial numbers and stability serve as a counterweight that diminishes institutional alterations.
This is precisely why Trump and Musk’s actions bear such weight. The further AI-driven decision-making is embedded within government, the more straightforward change will become. If human employees are broadly substituted with AI, executives will possess unilateral power to immediately modify the actions of the government, significantly increasing the stakes for transfer of power within a democracy. Trump’s unparalleled overhaul of the civil service might represent the final occasion a president needs to replace human workers to dictate new governmental functions. Future leaders may execute this at the touch of a button.
To clarify, the deployment of AI by the executive branch doesn’t necessarily have to be catastrophic. In theory, it could enable new leadership to promptly fulfill the desires of its electorate. However, this could turn disastrous if placed in the hands of an authoritarian figure. AI systems centralize authority at the top, potentially granting an executive the ability to implement change over sprawling bureaucracies instantly. Dismissing and replacing tens of thousands of human bureaucrats is an enormous task. Swapping one AI for another, or altering the regulations under which those AIs operate, would be significantly easier.
Social-welfare initiatives, if automated using AI, could be redirected to systematically favor one group and disadvantage another with a mere alteration in directives. Immigration enforcement agencies might efficiently prioritize individuals for investigation and detention with one command. Regulatory enforcement agencies monitoring corporate misconduct could whimsically shift their focus to or away from any particular company.
Even if Congress were assiduously trying to resist Trump and Musk, or any future president looking to override the intentions of the legislature, the ultimate authority to command AI agents would facilitate the circumvention of legislative intent. AI has the potential to undermine representative governance. Enacted law is never fully determinative of governmental actions—there is always room for maneuvering by presidents, appointed officials, and civil servants to apply their judgment. Whether intentional or inadvertent, whether generously or unfavorably, each of these actors exercises discretion. In human frameworks, that discretion is widely allocated among various individuals—people who, particularly in the case of career civil servants, typically endure longer than presidencies.
Currently, the AI landscape is dominated by a small number of corporations that dictate the design of the most widely utilized AI models, the data on which they are trained, and the instructions they adhere to. As their operations are mainly secretive and unaccountable to the public interest, these tech firms can modify the biases in AI systems—either generally or with specific governmental applications in mind—in ways that remain hidden from the public eye. Additionally, these private entities face potential coercion from political leaders and are motivated to sway in their favor. Musk himself established and financed xAI, now among the largest AI laboratories globally, with an explicit ideological agenda to produce anti-“woke” AI and influence the broader AI sector in a similar manner.
However, there exists an alternative trajectory for AI’s influence on government. AI development may occur within transparent and accountable public institutions alongside its continued evolution by Big Tech. Applications of AI in democratic governance could aim to enhance the services provided to public servants and the communities they assist by, for instance, facilitating non-English speakers’ access to government resources, streamlining ministerial processes such as handling routine applications, and alleviating backlogs, or aiding constituents in voicing their opinions on policies considered by their representatives. Such AI implementations should be undertaken progressively and meticulously, with public oversight throughout their design, execution, and monitoring to prevent unacceptable biases and harm.
Governments worldwide are illustrating how this can be accomplished, although it is still in the early stages. Taiwan has taken pioneering steps in utilizing AI models to foster deliberative democracy on an unprecedented scale. Singapore has emerged as a leader in the creation of public AI models, constructed with transparency and designed with public service applications in focus. Canada has demonstrated the importance of disclosure and public participation in deliberating AI usage in governance. Even if you lack trust in the current White House to emulate these models, U.S. states—which engage much more closely and have greater influence over the daily lives of Americans than the federal administration—could take the lead in responsible AI development and implementation.
As political theorist David Runciman has stated, AI represents yet another in a long succession of artificial “machines” employed to regulate human life and actions, similar to the corporations and states of the past. AI does not supplant those prior institutions, but alters their functioning. As the Trump administration cultivates closer ties with Big Tech and AI developers, we must acknowledge the potential of that collaboration to shape the future of democratic governance—and take measures to ensure that it does not empower future authoritarian regimes.
This essay was crafted in collaboration with Nathan E. Sanders and originally published in The Atlantic.