AI Isn’t Going to Cut Government Bureaucracy — It’s Going to Vastly Worsen It

In recent weeks, hundreds of foreign students in the U.S. have received emails from the Department of State informing them that they must leave the country. This “catch and revoke” program is being used to cancel the visas of students who have participated in forms of pro-Palestinian activism the government doesn’t approve of, which may include merely reading or posting certain kinds of content on social media. Thus, in little over a decade, we have gone from social media being touted as a weapon to topple dictators (recall the “Twitter revolutions” of the Arab Spring in the early 2010s) to a tool of mass surveillance that authoritarian governments can use to silence dissent. This time, however, the attack on free speech and the rule of law is aided by a new type of zealous bureaucrat: artificial intelligence.

While artificial general intelligence (AGI) remains a fantasy peddled by the right-wing tech oligarchy, we are now witnessing the creation of what I would call artificial bureaucratic intelligence (ABI). ABI is emerging as an autonomous agent dedicated to the rigid and mindless implementation of administrative procedure. As thousands of government workers are dismissed from their jobs, it appears that the plan is to replace them with some form of ABI, which will act as the interface between government (or what’s left of it) and the public. ABI is a “killer app,” because bureaucracy is the one area where failure, inefficiency and arbitrariness are features, not bugs. In 1979, Sen. Eugene McCarthy said that “an efficient bureaucracy is the greatest threat to liberty,” and ABI is poised to become the most effective bureaucrat ever. That’s because an AI bureaucrat won’t be distracted by a desire to flex its power or line its pockets — its sole purpose will be to follow instructions, no matter how disastrous the results.

We are already seeing hints of how this could work. In addition to catch and revoke, ABI implementations have been announced to identify government workers who can be fired and immigrants who can be deported (thanks to Donald Trump’s activation of the Alien Enemies Act). LGBTQ+ people and women seeking information about abortions have also been identified as targets of high tech surveillance. ABI might be initially used to go after populations unjustly considered to be a threat to national security. But as history has shown us, the definition of who or what constitutes a threat can be easily expanded to include anyone, from unionized public and private sector workers, to teachers, Muslims, BLM activists or people on welfare. ABI will act as judge, jury and executioner, carrying out a fascist agenda against anyone who dares question authority.

With the help of Big Tech, ABI could be instituted rather rapidly, since it is clear that the Trump administration is not necessarily interested in accomplishing things through democratic means. The most important thing at the moment seems to be setting a precedent, making ABI seem like a normal impending reality.

It is thus not surprising to see how few checks and balances seem to be in place to prevent the emergence of this reality. ABI is the result of decades of public sector defunding and neoliberal deregulation by both Republicans and Democrats. Tech companies have benefitted from this deregulation, but since their stocks have fallen sharply since Trump assumed power, they are now desperate for opportunities to monetize their overvalued AI technologies, and ABI represents just that.

Still, for all of this to work, one thing is needed, the one thing ABI cannot function without: our data. And since ABI needs lots of it to work, we must follow the data (not just the money) to understand why we can expect colonialist data policies to continue to be an essential part of this system.

In one corner we have the U.S. government, which ostensibly faces legal limitations on what data it can collect and how to use it. In addition to spying on us, the government has figured out it can simply buy our data from data brokers. This includes a wealth of information about practically every aspect of our lives, including health, financial, legal, education, employment and consumer data. It also includes all the information about our social activities and networks that is collected by social media companies.

In the other corner we have corporations, which are legally obstructed from accessing sensitive data that we entrust only the government to keep. Sure, corporations can access that data if they are doing contract work for the government, but there are usually lots of safeguards involved. However, a new solution has recently been attempted: Why not let the richest oligarch in the world simply take control of government data, in the name of efficiency? As we have seen in the last few weeks, Elon Musk’s so-called “Department of Government Efficiency” has been engaged in a digital coup that involves the capture of entire government data systems.

This unholy alliance between government and corporations points to what ABI is on the verge of becoming: a social credit system created by Big Tech (Alphabet, Amazon, Meta, Microsoft etc.) and sinister tech (Palantir, Clearview AI etc.), fed with our data, and used by the government to carry out its far right agenda. CEOs, pundits and academics used to signal their virtue by wagging their fingers at China and proclaim that at least we in the West were not trying to build a social credit system, a unified national mechanism for treating people according to the data that is collected from them. It now seems the U.S. is on its way to developing such a system overnight.

Is this truly a new development? Data has been an instrument of power and management for a long time. And yet, how computers make decisions about our lives is about to change fundamentally. Before ABI, algorithms were given specific criteria, and humans would then have to evaluate the results. For instance, an algorithm could be designed to determine whether insurance claims should be processed or denied, or to help pick targets in a military weapons system. Bias and abuse could be present in the system, certainly, but at least the logic of the algorithm could be examined.

ABI takes things a step further. There is no algorithm, no criteria. ABI can simply be given prompts (“identify pro-Hamas students”) and told how to apply a policy (“generate deportation cases for those students”). Accuracy — which is not AI’s forte — is not the goal. Its absence is in fact part of a strategy of state terrorism, because even the blameless live in fear of being singled out. A technology prone to hallucinations is considered deficient in applications that require small error margins and social context, but it is welcomed in situations meant to create chaos, persecution and unaccountability.

Lest this all sound rather dystopian, let’s keep another thing in mind: Automation and decision-making systems are already busy at work, directly intervening in the lives of U.S. residents, particularly the most vulnerable. According to TechTonic Justice, all 92 million low-income people in the U.S. are already experiencing some form of algorithmic or AI control over their lives. This means that, as we speak, these systems already have a say in determining how low-income people access disability and Social Security benefits, unemployment insurance, child welfare, supplemental nutrition and Medicaid services. These technologies already control how low-income people encounter services in education, language support, domestic violence and housing. And AI systems are already making decisions about how low-income people interact with the legal, immigration, tax enforcement and voting systems. In many cases, the automated systems that claim to reduce fraud end up unjustly punishing these vulnerable populations.

With ABI, these interactions will be extended to all corners of society, as public sector agencies become increasingly governed by the principles of what sociologist George Ritzer called McDonaldization: efficiency, calculability, predictability and control. Guided by these principles, ABI will create a public service technocracy where most of us will experience not just an erosion of access, but the continuation of a system that uses our data to actively exploit us for profit and social control. As we continue to watch the Trump administration dismantle crucial public institutions, and as we work to not just take back but to remake our democracy, we must keep in mind that ABI systems designed in the name of efficiency, calculability, predictability and control are, and will always be, anti-democratic and inhumane.