Just after the election, President Trump and his allies promised to cut government spending through euphemistic appeals to “efficiency,” suggesting that artificial intelligence could do this effectively.
Now in office, Trump has empowered Elon Musk to helm the project of “AI-First” governance, which includes commandeering sensitive government data, such as U.S. Treasury payments, Medicaid and Medicare systems and the personal records of millions of Americans. Musk’s Department of Government Efficiency conspicuously began its rampage after courts halted Trump’s illegal scheme to freeze congressionally-approved funds to programs conservatives dislike.
The tactic exposes the administration’s shell game of executing the president’s agenda by using AI to term anything that he opposes as “waste, fraud or abuse.”
The term “fraud” carries specific legal definitions beyond something that makes Trump or Musk mad — it involves intentional deception to obtain government benefits or contracts. AI is terrible at making these sorts of determinations — leading, for example, to false fraud allegations against nearly 40,000 people in Michigan who collected unemployment benefits, as well as months-long delays of early pandemic-era benefits to millions of desperate people.
These and countless other examples show AI for what it is: Not a rational way to make unbiased decisions, but rather a commercial product fueled by an unprecedented investment hype cycle that has been hawked by tech companies, often to organizations lacking the expertise needed to interrogate the technology’s promise.
But no autocrat gains power by heeding pesky laws or facts. Musk will simply program the AI to define governmental waste or fraud in terms of Trump’s stated enemies: civil rights, diversity, transgender people, reproductive health care (including abortion), global warming, science, journalism, immigrants and low-wage workers.
The AI will scour vast Treasury Department records looking for payments associated with these enemies and flag them. Similar efforts are already underway at the Department of Education and the National Science Foundation, which, for example, has been ordered to dig through already-approved grants for any mention of “women” or “people of color.”
Next, things may turn uglier. Trump can again attempt to freeze spending for some of these programs by hewing more closely to the Impoundment Control Act, which allows the president to temporarily freeze payments while seeking congressional approval for a permanent halt. The administration will likely justify freezes by stating that AI — a supposedly “objective” arbiter — has identified the funds as fraudulent or wasteful.
Even if Congress ultimately rejects the freezes and the money is eventually paid out, the programs and their staff will have already been destabilized. Moreover, Trump will have advanced the groundwork for impending budget negotiations by labeling disfavored programs as fraudulent, giving his congressional allies targets to eliminate and political cover for doing so.
At the same time, expect Musk to build an elaborate mythology around AI as a way to revolutionize government. His acolytes already claim, without evidence, that AI can automate complex processes and reduce staff. To this end, the government recently announced a massive purchase of AI services from OpenAI, the company that makes ChatGPT.
Of course, reduced government staffing due to automated work means a loss of expertise and capacity needed to serve the public. This in turn degrades public trust and justifies more cuts. Doing so to enrich a company like OpenAI that is facing financial challenges reeks of corruption, which makes the system feel rigged against everyday Americans. Trump may slash Medicaid and SNAP benefits for people struggling, only to give handouts to his Big Tech pals.
AI will be foundational to Trump’s policy project in other ways, too, such as by facilitating mass deportations, identifying protesters for retaliation or advancing militarism. Republican-controlled states will find AI useful to review state funding for disfavored programs, hunt down patients seeking abortions and their doctors, cut health care benefits to eligible people or force disabled people to go without caregiving assistance.
With Trump all-in on AI, the tech sector is ready to support him. Big Tech CEOs contributed to and attended Trump’s inauguration. Meta (the parent company of Facebook and Instagram) ended fact-checking practices that slowed the spread of the kind of misinformation Trump and his allies espouse. Google removed its restrictions against building AI for weapons and surveillance. And Google, Amazon and Meta have abandoned diversity commitments in hiring in order to appease Trump.
Still, for all its momentum, government-by-AI isn’t inevitable. Advocates, including myself, have won many legal challenges under well-established laws touching on due process, access to public information and anti-discrimination. Unions representing federal workers are invoking privacy laws in suits to stop the flow of data that Trump’s AI needs to function. Even if lawsuits are difficult, they offer ways to stave off some tech-driven harms, slow AI’s implementation and highlight abuses.
Meanwhile, the American people do not trust AI. The administration’s current AI blitzkrieg may intensify this antipathy, as it will further threaten the 2.3 million federal employees already facing coerced resignations; make government benefits like Social Security and Medicaid less accessible to the roughly 73 million people who rely on them; and potentially disrupt other key areas of people’s lives, such as education, work and housing.
Americans are similarly skeptical of Big Tech, and with good reason: the industry’s perceived greed, increasing concentration of corporate power, documented worker mistreatment, the nuisances of data centers, intensive environmental devastation and ongoing erosion of privacy.
Taken together, conditions seem ripe for a mass movement of people harmed by these technologies to organize against their proliferation and use. In the end, Trump’s turn towards Big Tech to build him an AI crown may backfire.
Kevin De Liban is the founder and president of TechTonic Justice, a nonprofit organization advocating for the people AI leaves behind. He was previously a legal aid attorney for 12 years.
