Five years ago, at the Social Security office in Salem, I discovered that I wasn’t a US citizen — according to the agency’s database. Although I had been naturalized as a teenager, no one informed Social Security. It was a maddening but ultimately harmless bureaucratic error, one easily fixed by producing my passport.
But today, the mistake might have had dire consequences. With the Trump administration vowing to deport millions of undocumented people, Immigration and Customs Enforcement collects vast amounts of government data to feed the AI platforms the agency uses. This year alone, billions of dollars will be spent on tech companies that aim to make deportation more efficient, even though AI is prone to error and difficult to challenge, leaving even legal residents at risk. And that leads to an alarming question: Who controls the information infrastructures the country uses to make decisions about people’s lives?
Governments have been tracking people as far back as the earliest civilizations in Mesopotamia, so we shouldn’t be surprised it’s happening to us too. During the 19th century, federal officials mined ship manifests to document names, ages, occupations, and destinations of those just arriving in Boston — an understandable approach to increased immigration. But when the government began restricting entry under the Chinese Exclusion Act of 1882, it needed new ways to document identity. Chinese immigrants were forced to carry certificates verifying they were here legally.
Later, the Alien Registration Act of 1940 required noncitizens to submit their fingerprints; biometric data made them easier to track. Now, as masked agents sweep through US cities, innovation continues. ICE reaches into any database it can — Social Security, municipal license-plate readers, registries of motor vehicles, and others to fuel AI platforms.
It’s even pursuing data long believed to be confidential, such as IRS filings. Whether we’re citizens or not, our names, addresses, and tax returns are only supposed to be shared in exceptional circumstances, such as federal criminal probes. Yet, by June 2025, the IRS had handed over 47,000 taxpayer records to ICE for “criminal investigations.” Two federal judges have since halted an agreement between the two agencies, ruling that IRS data should only be disclosed for discrete criminal cases, not in bulk.
ICE needs mountains of data for its decision-making systems, such as the $30 million AI surveillance platform ImmigrationOS, which Palantir Technologies is developing. The software will predict who might have overstayed their visas, be part of a criminal organization, or be violent criminals, then it generates lists of people to deport. ICE will be able to use it to optimize every part of the immigration enforcement process, from identifying people to detain to ensuring they have left the country.
Amnesty International has warned that such technologies, which make predictions about people’s behavior using statistics, have high margins of error. (A similar platform used by the Israel Defense Forces, built with Palantir infrastructure, incorrectly flags 10 percent of people analyzed as members of Hamas.) There is also the risk of automation bias, where the humans deciding whether to pursue a target will ignore contradictory information or their own conscience because they trust the technology.
AI in immigration enforcement is just one example of how the US government depends on public-private partnerships to perform one of its oldest functions: turning information into power. Together, tech corporations form a dispersed, privately governed network that builds data infrastructure and holds tremendous control over public life: a data empire, ruled by CEOs.
Amazon Web Services hosts multiple federal systems, including welfare databases and classified intelligence workflows. The Department of Defense contracts with Microsoft Azure for data strategy, modernization, and cybersecurity. Google manages some data processing and web hosting for the Department of Agriculture.
Collectively, tech companies represent the largest threat to the future of democracy. Decisions, including who receives benefits, who’s a threat to national security, or which farmers receive loans, are being made with systems created by companies shielded from public scrutiny. Outsourcing the vital technologies needed to govern cedes democratic accountability to a largely unregulated tech industry motivated by the bottom line.
In the case of immigration enforcement, the price tag is growing. The Secure America Act, passed in June, earmarked $5 billion for ICE and Customs and Border Protection to spend on AI, according to a recent report by immigrant rights group Mijente along with Just Futures Law and the Surveillance Resistance Lab. From 2024 to 2025, contracts with Clearview AI for facial recognition alone had more than tripled.
But around the country, people are pushing back. Last year, Delaware lawmakers acknowledged residents’ concerns and banned local law enforcement from cooperating with ICE. Last month, the City of Salem abandoned its contract with Flock, an automated license plate reader camera system — a result of grass-roots activism by a coalition of organizations and residents. Salem is among more than 80 cities rejecting Flock due to concerns over mass surveillance, especially of immigrants. Earlier this year, people in Maine filed a class action suit alleging the Department of Homeland Security is using biometric and license plate data to intimidate residents observing immigration enforcement operations.
These activists remind us that we have a say in the technologies that govern our lives and the way our data are used. Ensuring that AI strengthens democracy instead of undermining it depends on our willingness to demand transparency, accountability, and public oversight from the government and the companies on which it depends.
We have already seen permanent residents and citizens detained and deported. How many more life-changing mistakes will be made? What other decisions will AI make for the country? And how many more of our tax dollars will be spent on technologies that make the US government dependent on Silicon Valley without regulation to protect us?
Roopika Risam is the author of Data Empire and a professor at Dartmouth College. Send comments to magazine@globe.com.