Who Will AI Serve?
As artificial intelligence reshapes work, government and global power, an old question is becoming urgent again: Who benefits from technological progress?
Artificial intelligence spent much of 2026 racing forward. New models became more capable. Companies expanded their investments. Governments explored new uses for AI. Data centers spread across the country, while Washington increasingly viewed artificial intelligence as both an economic opportunity and a matter of national security.
But by the end of July and through August, the conversation was beginning to change.
The question was no longer simply, “What can AI do?”
Increasingly, it became, “Who will AI serve?”
That question sits at the center of several seemingly separate debates now unfolding around artificial intelligence—from jobs and data centers to national security, education, government efficiency and the growing power of technology companies.
It also received an unexpected voice from the Vatican.
Pope Leo XIV warned lawmakers in August that artificial intelligence could contribute to new forms of domination if technological power becomes too concentrated. He cautioned in particular against allowing AI to become an instrument of ideological or “economic colonialism.”
The phrase is striking, but the concern behind it is straightforward.
Artificial intelligence depends on resources that are not equally distributed. Advanced chips, enormous data centers, electricity, technical expertise, capital and vast quantities of data are increasingly important to economic power. Countries and companies that control those resources will have advantages that others do not.
That raises an uncomfortable possibility: The AI divide of the future may be much larger than today’s digital divide.
Some countries may build and control intelligent systems. Others may simply rent access to them.
The same divide could develop within countries. Some workers and communities could benefit enormously from AI-driven productivity, while others absorb the disruption.
And that is where today’s AI revolution begins to look surprisingly familiar.
We Have Been Here Before
Carl Benedikt Frey explores that history in The Technology Trap: Capital, Labor, and Power in the Age of Automation.
Frey’s lesson is not that technology is bad. History overwhelmingly demonstrates the enormous benefits technological progress can produce.
The problem is the transition.
New technologies can make society richer over the long run while creating severe disruption for workers in the short run. The benefits also do not necessarily reach everyone at the same time.
Consider the Luddites.
Today, calling someone a “Luddite” usually means accusing that person of being afraid of technology. The original Luddites had a more complicated concern.
Machines were changing their economic value.
Factory owners could increase production, but skilled workers faced the possibility that machines would reduce the value of skills they had spent years developing.
Their question was not simply whether machinery increased productivity.
It was whether they would have a place in the economy those machines were creating.
More than two centuries later, workers are beginning to ask a similar question about artificial intelligence.
This Time, the Office Is Part of the Factory
Earlier waves of automation transformed farms and factories. AI is moving directly into offices.
Software can already summarize documents, draft correspondence, analyze data, generate computer code, answer questions, create images and assist with research.
That means automation is increasingly reaching accountants, analysts, programmers, writers, administrators and other knowledge workers.
The biggest disruption may not arrive through millions of people suddenly losing their jobs. It could begin more quietly—with fewer entry-level positions.
That presents a problem that deserves more attention.
Entry-level jobs are not just inexpensive labor. They are how people learn professions.
The experienced analyst was once the junior analyst. The manager was once the new employee. The attorney once handled basic research. The inspector once accompanied experienced inspectors. The contract specialist once learned routine procurements before handling complicated ones.
Those assignments build judgment.
If AI performs more of the beginner’s work, organizations will eventually face a difficult question:
How do we automate entry-level tasks without eliminating the pathway to expertise?
Government should be particularly concerned.
Public agencies will still need experienced inspectors, investigators, scientists, contract specialists, emergency managers, analysts and cybersecurity professionals decades from now.
AI might help those employees become dramatically more productive. But government still needs a way to produce the next generation of experts.
The goal should not simply be replacing people.
It should be making people more capable.
There Is Another Side to the Argument
Yet fear of disruption creates another danger.
Alexander C. Karp and Nicholas W. Zamiska make that case in The Technological Republic: Hard Power, Soft Belief, and the Future of the West.
Their concern is almost the reverse of Frey’s.
Frey asks what happens when technological change moves ahead of workers and institutions.
Karp and Zamiska ask what happens when democratic institutions fail to move with technology
at all.
Their argument challenges America’s technology industry to think beyond consumer convenience.
Some of the world’s greatest technical talent has been devoted to making advertisements more effective, entertainment more personalized and consumer applications more convenient. Those products can certainly have value.
But a country has bigger problems.
National security. Aging infrastructure. Scientific discovery. Cybersecurity. Energy. Public health.
Transportation. Emergency response. Government modernization.
Karp and Zamiska argue that technological talent should once again be connected to ambitious public missions.
That idea has deep roots in American history.
Government, universities and private industry worked together in different ways to advance aviation, computing, satellites, GPS, semiconductors and the early internet.
The lesson is not that government must control innovation.
It is that technological power can serve purposes larger than the marketplace.
AI now presents the same choice.
Washington and Silicon Valley Need Each Other
The United States therefore faces two risks at once.
Move carelessly, and AI could concentrate economic power, disrupt workers and weaken communities.
Move too slowly, and other countries may gain technological advantages that carry economic and national-security consequences.
There is no easy choice between “innovation” and “regulation.”
America needs both innovation and capable institutions.
Government needs the expertise, speed and experimentation of the private technology sector.
Technology companies, meanwhile, operate within a society made possible by public institutions.
Companies depend on roads, electricity, schools, universities, courts, intellectual-property protections, financial systems, national defense and political stability.
Neither side can fully replace the other.
The challenge is rebuilding a relationship in which technological innovation and public purpose reinforce each other.
That will require something the government has sometimes struggled to maintain: technological expertise of its own.
Government Cannot Outsource Understanding
As federal, state and local agencies adopt AI, public administrators will make thousands of decisions that determine how the technology actually affects citizens.
Contract specialists will decide what systems agencies purchase.
Program analysts will determine whether they improve performance.
Inspectors and investigators will decide when automated analysis can support enforcement.
Human-resources offices will confront workforce changes.
Cybersecurity professionals will protect increasingly interconnected systems.
Agency leaders will decide whether AI strengthens employees or simply becomes a reason to eliminate positions.
These decisions may sound administrative.
Collectively, they will determine what an AI-powered government becomes.
Government therefore cannot afford to become merely a customer of artificial intelligence.
Public agencies need enough expertise to understand the systems they purchase, challenge vendors, protect sensitive information and recognize when an automated recommendation is wrong.
Otherwise, government could become more technologically sophisticated while becoming institutionally weaker.
That would be a dangerous trade.
Efficiency Is Not the Same as Capacity
Imagine an agency introduces an AI system that can perform work previously completed by hundreds of employees.
The immediate savings might look impressive.
But look ten years ahead.
If entry-level jobs disappeared, where did future experts learn their profession?
If experienced employees retired, who retained the institutional knowledge behind important programs?
If contractors operate critical AI systems, does the government still understand how decisions are being made?
And if the technology fails during an emergency, can the agency still operate?
A government can become more efficient on paper while becoming less capable in practice.
That distinction should shape how public agencies measure AI.
The best system may not be the one that eliminates the most positions.
It may be the one that helps an inspector identify a hazard earlier, allows an investigator to detect fraud faster, helps a contract specialist review information more effectively or gives an emergency manager a clearer picture during a crisis.
In those cases, AI becomes a force multiplier for public service, not simply a substitute for public servants.
AI Is Not Just in the Cloud
There is another side of artificial intelligence that is becoming increasingly difficult to ignore.
AI may appear digital, but its infrastructure is remarkably physical.
Data centers need land.
They need enormous amounts of electricity.
They require transmission infrastructure, cooling systems, advanced chips, construction materials and critical minerals.
That means the AI boom will not remain confined to Silicon Valley.
Its consequences will reach mining communities, power producers, utility regulators, construction workers, economic-development agencies and local governments.
Some communities will host data centers. Others will produce the electricity powering them.
Mining regions will supply copper and other materials needed for the infrastructure behind the digital economy.
Every technological revolution creates geography.
Railroads created transportation hubs. Manufacturing created industrial cities. Automobiles reshaped suburbs. The internet helped create new centers of economic power.
AI will do the same.
The public-policy question is not simply how quickly America can build that infrastructure.
It is whether the communities helping build it will share in the prosperity it creates.
That brings us back to Frey—and to Pope Leo.
Teaching People to Question the Machine
The transition also begins in schools.
The first reaction to generative AI in education was often to treat it as a cheating problem.
That approach is becoming harder to maintain.
Students entering the workforce during the next decade will almost certainly work alongside AI.
They therefore need more than instructions on how to use it.
They need AI literacy.
Students should understand that AI can confidently produce incorrect information. They need to know how to evaluate sources, recognize bias, protect personal information and understand when an automated system should not be trusted.
Most importantly, they must learn how to question it.
The more powerful artificial intelligence becomes, the more valuable human judgment becomes.
Three Warnings, One Question
Pope Leo, Frey, and Karp and Zamiska approach technology from very different directions.
Yet their ideas meet at the same point.
Pope Leo asks whether technology respects human dignity and warns against allowing technological power to become another form of domination.
Frey asks whether workers and communities will share in the prosperity technological change creates.
Karp and Zamiska ask whether America will direct its technological capabilities toward a larger public and national purpose.
Those questions should not compete with one another.
They belong together.
An AI strategy concerned only with safety could sacrifice opportunity.
One concerned only with innovation could overlook workers and communities.
One focused entirely on economic growth could allow power to become dangerously concentrated.
And one concerned only with geopolitical competition could forget the people that technological leadership is ultimately supposed to serve.
The challenge for public administration is balancing all of them.
The View From 2076
That challenge carries special meaning as the United States marks its 250th anniversary.
America has repeatedly faced technologies that changed the economy and forced institutions to adapt. Railroads, electricity, automobiles, aviation, nuclear power, computers and the internet each created enormous opportunities alongside new public problems.
Artificial intelligence may be the next great test.
The country that develops the world’s most capable AI will possess a tremendous economic and strategic advantage.
But that alone will not determine whether the AI revolution succeeds.
The stronger test will be what happens to the people living through it.
Can workers participate in the prosperity automation creates?
Can young people still move from beginners to experts?
Can communities supporting AI infrastructure share in its benefits?
Can Government use AI without surrendering accountability?
Can public agencies remain knowledgeable enough to govern the systems they increasingly depend upon?
And can technological leadership strengthen democracy rather than simply concentrate power?
Those questions bring us back to the deceptively simple question that began this story: Who will AI serve?
Pope Leo reminds us that the answer must begin with people.
Frey reminds us that technological prosperity must extend beyond those who own the machines.
Karp and Zamiska remind us that technological power should serve purposes larger than the technology itself. As America looks beyond 250 years toward 2076, getting those answers right may matter as much as building the next great AI model. The technology will help determine what is possible. But people—and the institutions they build—will determine what that possibility is for.
