Submission by Tom DeWeese and Eric Meder, American Policy Center

Artificial intelligence is rapidly transforming American life, while massive AI data centers are being built across the country to provide the computing power behind the technology. Supporters say these facilities will create jobs, attract billions of dollars in investment and help the United States compete with China for global AI dominance. But as artificial intelligence becomes increasingly capable of collecting, analyzing and connecting enormous amounts of personal data, concerns are growing about privacy, government surveillance and individual freedom. Are AI data centers building the foundation for unprecedented prosperity, or creating the infrastructure for a surveillance system more powerful than anything Americans have experienced before?

“He that walketh with wise men shall be wise, but a companion of fools shall be destroyed.” Proverbs 13:20 (KJV)

President Donald Trump has embraced data centers as engines of investment, employment and American technological strength. When New York paused permits for new hyperscale facilities in July, Trump called data centers “cash cows” capable of producing lower taxes and record-setting jobs. He blamed “Radical Left Democrats” for resistance that, he warned, could help China overtake the United States.

That argument has enormous appeal in Washington and in communities promised billions of dollars in construction. The people who would live beside these projects have questions that cannot be waved away as partisan obstruction. Local opposition has crossed political lines because the consequences are intensely local: who pays for new power infrastructure, how much water a facility will use, what tax concessions a developer will receive and whether residents retain a meaningful voice in land-use decisions.

The demand for power is real. Data centers consumed about 4.4% of all U.S. electricity in 2023, according to a Department of Energy report. The department projected that their share could rise to between 6.7% and 12% by 2028. That national figure will be felt unevenly, with the heaviest pressure falling on the grids and communities where large clusters of facilities are built.

The debate has already moved beyond zoning hearings. Legislators are deciding whether utilities may pass infrastructure costs to households and small businesses, whether states may override local restrictions and what information developers must disclose before a project is approved. Some states have protected local authority while others have considered preemption, leaving Americans with very different rights depending on where they live.

President Trump is right that the United States cannot abandon advanced computing to China. National competition, however, does not require local governments to accept every proposal on a developer’s terms. Prosperity claims should be tested in public, against binding commitments on electricity, water, taxes, noise, emergency generators, land use and permanent employment. Officials who dismiss residents instead of answering them deepen the suspicion surrounding these projects, as Katy Christian Magazine has reported in its examination of the tactics used to overcome data center opposition.

Computing Power Is Only Part of the Story

An AI data center supplies computational muscle. Inside are servers, networking equipment, power systems and cooling infrastructure that allow companies and government agencies to train models, store information and analyze enormous datasets. The server racks have no inherent ability to read text messages or track a driver; those records must first come from phones, apps, connected vehicles, cameras, financial institutions, websites or government databases.

Once those streams are available, powerful computing systems can search them rapidly, combine information gathered for different purposes and draw inferences about a person’s movements, relationships or behavior. Collected data, computing power and an institution authorized to act upon the results: That is the combination that turns infrastructure into surveillance.

Technology companies also do extensive business with federal agencies. The government can obtain information through warrants, subpoenas, court orders, intelligence authorities and commercial contracts. Section 702 of the Foreign Intelligence Surveillance Act, for example, permits surveillance targeting non-Americans reasonably believed to be outside the United States and requires electronic communications providers to assist when lawfully directed. The program is narrower than a license to inspect any American’s device at will, yet the Privacy and Civil Liberties Oversight Board has repeatedly examined the consequences for Americans whose communications are collected or queried through the system.

Oracle co-founder Larry Ellison supplied an unusually candid picture of where the industry may be heading. During the company’s 2024 financial analyst meeting, he discussed police body cameras and vehicle cameras transmitting video for continuous AI monitoring. “Citizens will be on their best behavior,” he said, because authorities would be “constantly recording and reporting” what was happening.

Ellison presented the system as a way to deter misconduct by citizens and police alike. His confidence in pervasive monitoring raises a question that cannot be answered with a promise of public safety: Who watches the resulting footage, who may search it, how long is it retained and what happens when the system is wrong or abused?

When a Computer Remembers the Screen

Microsoft’s Recall feature provides a concrete example of both the appeal and the risk. Recall is available on certain Windows 11 Copilot+ PCs and can save periodic snapshots of activity on the computer, allowing a user to search for something previously seen on the screen.

The feature does not take screenshots from a person’s phone or smart television. Following an intense privacy backlash, Microsoft made Recall opt-in and added safeguards. The company says snapshots remain encrypted on the computer, are not sent to Microsoft or used to train its AI models and can be opened only after the user authenticates through Windows Hello. Users may pause Recall, exclude websites and apps, set storage limits or delete the saved material. Microsoft describes those controls in its security and privacy architecture.

Even with those protections, Recall creates a valuable archive of a person’s activity on one machine. A compromised account, malicious software or unauthorized physical access could expose private conversations, documents and browsing history. People should be told clearly what is being saved and should never have to hunt through menus to stop it.

Identity, location and payment systems pose parallel problems, which Katy Christian Magazine examined in “AI and the Cashless Society: Freedom or Slavery?” As institutions make more decisions from connected data, consent, correction procedures, access logs and human review become essential protections.

Palantir and the Power to Connect Records

Palantir’s Gotham platform is used by military, intelligence and law enforcement organizations to work with large, sensitive collections of information. Palantir describes Gotham as software that helps analysts and operators integrate data, investigate relationships, track events and make operational decisions. Its published interface includes observations, maps, targets and high-priority target lists.

Calling every use of Gotham “pre-crime” overstates what the software does. Its capabilities still warrant close scrutiny. A system that brings records from separate databases into one environment can reveal associations that would have been difficult to see manually. It can also magnify bad data, place innocent conduct in a suspicious context and distribute an error across agencies before the affected person knows the record exists.

Public oversight has to reach the details: which databases feed the system, the legal authority covering each one, whether an analyst must document a reason for searching a name and whether a result is treated as an investigative lead or as evidence. Citizens also need a way to correct inaccurate information, while independent auditors need records showing who searched the system and why. The answers depend on the agency, the contract and the configuration. “Trust the software” and “trust the government” are no substitute for enforceable limits.

Flock Cameras Are Already in Katy

Katy drivers already pass one piece of this surveillance network along local roads.

Flock Safety’s fixed cameras are automated license plate readers, not facial-recognition cameras. They photograph vehicles and record details such as the license plate, location, time and visible vehicle characteristics. Police can search that information during an investigation and receive alerts when a detected plate matches a list entered into the system. Flock says its plate-reading system does not use facial recognition.

As of August, the city of Katy had expanded from 22 to 40 Flock license plate readers, according to records reviewed by Katy Christian Magazine. The cameras may help locate a stolen vehicle or identify a car connected to a serious crime. They also build a searchable history of ordinary drivers who were never suspected of one.

The network’s reach is determined by retention periods, search rules and agreements that permit other agencies to use the data. Those policies decide how much of a person’s travel can be reconstructed long after the car passed a camera. Every search should require a legitimate purpose and produce an audit record. Agencies should disclose how often the cameras contribute to arrests, how frequently alerts are wrong, which outside agencies can search local records and whether officers have been disciplined for misuse.

When officials invoke public safety, the public is entitled to evidence and limits. In practice, the threat AI can pose to individual freedom will be decided through ordinary policy choices: what is collected, who receives access and whether the public can see what government is doing.

AI and the Power to Kill

National security is among the strongest arguments for American investment in AI. It is also the field in which an automated mistake can carry the greatest cost.

AI systems do not possess a conscience, fear death or bear moral responsibility for a command. They calculate responses from training, instructions and available data. Their speed may help commanders sift intelligence or detect threats, but speed can become dangerous when the evidence is incomplete and the time for human judgment is compressed.

Researchers tested five language models in simulated military and diplomatic crises and found that all displayed forms of escalation and difficult-to-predict behavior. The models tended to produce arms-race dynamics; in rare cases, they chose to deploy nuclear weapons. The study did not find that AI always sacrifices human life, and a simulation cannot tell us exactly how a system would behave in a real command structure. It did show why no language model should receive autonomous authority over lethal or nuclear decisions.

Human control must mean more than placing a person at the end of a process and asking for a hurried approval. Decision-makers need time, contrary evidence, an understanding of the model’s limits and the authority to reject its recommendation. No company should be permitted to hide a military system’s errors behind trade secrecy when human lives are at stake.

A Warning President Trump Should Hear

President Trump should listen to the Americans asking what these facilities will cost their communities and what the technology they power will be used to do. They are homeowners, ratepayers, farmers, parents and local officials. Many support American technological leadership while refusing to surrender local control, personal privacy and constitutional protections as the price.

Proverbs 13:20 warns that the company a leader keeps can shape his judgment. An administration surrounded by industry promoters will hear about investment totals, construction schedules and the race with China. It also needs advisers willing to discuss surveillance, utility costs, water, property rights and the concentration of power.

AI can be a valuable tool. It can also give institutions the ability to monitor, classify and act upon people at a scale no previous government or corporation possessed. Data centers make that scale possible. Before approving more of them, public officials should require transparent agreements, protect local decision-making, establish firm limits on surveillance and preserve human authority over decisions involving liberty or life.

American strength has never rested on computing power alone. It rests on citizens who remain free to speak, own property, govern their communities and live without being cataloged at every turn. Any technology built in the name of national greatness must answer to those principles.

 

Feature photo: Anti-data center sign, Catboy69, CC BY 4.0 <https://creativecommons.org/licenses/by/4.0>, via Wikimedia Commons