The AI Backlash Is a Fight Over Who Owns the Upside

AI is currently crossing the line every major technology eventually crosses: it is becoming something people increasingly fight over in the public square.

For the last few years, AI anxiety mostly looked like scattered arguments: artists fighting training data, workers worrying about job loss, parents worrying about cheating in school, and critics as well as some tech leaders warning that models might become dangerous. Now, the fight against AI (and to an extent, the larger fight against big tech) has culminated in a new target: towns fighting data centers. Therefore, the scattered era of "little fights everywhere" is giving way to a larger, more institutional response.

In one corner, "The State" is preparing to treat parts of the backlash as a security problem. WIRED obtained 1,000+ pages of unpublished reports from DHS, the FBI, and what are called "fusion centers" (regional US intelligence hubs linking federal, state, and local police). These reports showed law enforcement tracking a new category of concern: “anti-technology extremists.”

In the other corner, "The Church" is preparing to define the moral stakes in an attempt to forge a more human-friendly path forward for the future. Pope Leo XIV published Magnifica Humanitas, a massive encyclical on safeguarding the human person in the age of AI. His frame is simple: humanity can build Babel, a tower of profit, control, and technological domination, or Jerusalem, a shared city where everyone contributes and benefits. Well, when you put it like that...

That is the surprising thing about reading a WIRED investigation next to Pope Leo XIV's new treatise on humanity's future in the AI age. They come from radically different institutions, but they both represent the first real institutional attempts to respond to the same new reality: AI has become real-world infrastructure, and real-world infrastructure creates real-world political reactions.

The deeper question at the heart of both of these attempts to react to the current moment is ownership of the future. People can accept some technological disruption when they believe they have a voice, a share of the gains, and a credible reason to think the future is about to get better for them. But they can revolt when the costs are primarily local, and the upside feels like an abstract, offshore, or already foregone conclusion that they're not a part of.

First up, the TL;DR

Across the country, the AI boom is getting a visible manifestation in the form of data centers: giant facilities full of servers that need land, electricity, water, backup power, substations, tax deals, zoning approvals, and neighbors willing to live near the hum. That turns AI from a software story into a local politics story. And as the saying goes, all politics is local...

This week, two institutions started saying the quiet part out loud.

The State (in the form of the US government) is preparing for anti-tech unrest. The Church (in the form of the Catholic Church) is warning against building a techno-corporate Babel.

And both are the first true institutional reactions to the larger AI boom: this AI stuff is really here to stay, so what are we going to do about?

Here's how The State is reacting:

• WIRED obtained 1,000+ pages of unpublished reports from DHS, FBI, and fusion centers showing law enforcement tracking a broad new category of risk: “anti-technology extremists.”
• The reporting connects the new focus to three pressures building at once: attacks on CEOs, data center protest movements, and fear that AI will replace jobs.
• A New York intelligence report warned that AI adoption could fuel “large-scale protests” that devolve into civil unrest and “anti-tech violent extremist activity,” especially in large cities.
• WIRED says the phrase “anti-tech violent extremism” does not otherwise appear in public DHS or FBI domestic extremism reports, making it a new and potentially elastic category.
• The story's core tension: some anti-tech violence is real, but the category can widen fast enough to sweep in peaceful data center critics, AI skeptics, and local residents fighting projects in their towns.
• One example WIRED flags: a SITE Intelligence report circulated a More Perfect Union video about a Georgia data center as a possible threat vector, even though the video did not advocate violence.

And of course, US President Trump had planned to sign an AI Executive Order last week, but scrapped it at the last second due to unnamed concerns, about which President Trump said he "didn't like certain aspects of it." You can read the full draft here.

Separately, here's how The Church is reacting:

• In Magnifica Humanitas, Pope Leo XIV frames AI as a civilizational choice between Babel and Jerusalem.
• • Babel is the path of uniformity, domination, self-sufficiency, and efficiency that sacrifices human dignity.
  • Jerusalem is the shared rebuilding project where everyone has a role.
• The Pope says technology can heal, connect, educate, and protect creation, but it can also divide, exclude, and create new injustice. His key line: technology is never neutral because it takes on the priorities of the people who design, fund, regulate, and use it.
• He draws a sharp line between AI and human intelligence. AI can imitate some human functions, but it does not have a body, experience, love, conscience, responsibility, suffering, or inner growth.
• He warns that AI systems touching employment, credit, public services, reputation, or opportunity can create new forms of exclusion when people cannot understand, challenge, or appeal automated decisions.
• He argues that data should carry social obligations because it comes from many contributors, which means it should be managed creatively as a common or shared good.
• His practical economic answer is surprisingly concrete: every introduction of automation and AI should include verifiable protections for employment, retraining, and worker participation.

That is the deeper story here. The State sees a security problem. The Church sees a legitimacy problem.

Both are responding to the same pressure: people are starting to feel that AI is being built around them, near them, and sometimes on top of them, without them.

Why this matters: Data centers are becoming AI's defining political fault line: data centers are where hidden tensions about jobs, power, water, wealth, health, and national competitiveness all bubble up like molten lava and break through to the surface.

• Gallup found that 70% of Americans oppose AI data centers in their local area, including 48% who strongly oppose them. Only about a quarter favor them.
• Gallup also found data centers are more locally unpopular than nuclear plants, with 71% opposing AI data centers nearby versus 53% opposing a nuclear plant nearby.
• Erin Brockovich's AI data center map tracks major AI-focused and hyperscale data centers running AI workloads, plus community-submitted reports about local impacts.
• According to Nieman Lab, Brockovich's map showed 33 operational data centers, 44 under construction, 27 proposed, and 2,716 community reports as of publication.
• The Brockovich site names six core community concerns: energy consumption, water use, e-waste, location risks, scalability and efficiency, and noise.
• At the same time, Goldman Sachs projects annual AI infrastructure spending will reach $800B by the end of 2026, covering chips, servers, data centers, and software.
• Goldman also raised its 2026 capex growth forecast from 6.5% to 7.8% because of the AI buildout, which helps explain why data centers are now both a neighborhood fight and a national growth story.

That is the uncomfortable tension: the same infrastructure that supports U.S. AI leadership, stock-market gains, retirement accounts, local tax bases, and future productivity is also raising fears about water, power bills, noise, air quality, land use, job displacement, and who actually gets paid from all this "boom."

Our take: Local communities are treating datacenters like a referendum on the AI boom, and we'd be shocked if this referendum doesn't play a part in the ballot box later this year. AI companies and policymakers have been treating data centers just like infrastructure, when they're really a Rorschach test that everyone imprints their own meaning onto.

This is where the Pope's Jerusalem metaphor comes into play: Jerusalem is technology built through shared responsibility. Babel is what happens when one language, one system, and one class of builders tells everyone else the tower is already decided.

The bigger test is whether the AI buildout gives ordinary people a reason to believe the future is being built with them, and that they own a piece of what comes next.

That could look like the following:

• Participation: towns, workers, parents, schools, utilities, and civic institutions need a real voice before the project is locked.
• Ownership: affected communities need a visible piece of the upside, whether through utility-bill credits, local tax guarantees, community benefit trusts, worker retraining funds, public AI dividends, or shared energy projects tied directly to the data center.

Now, let's dive into all of that with more detail, shall we?

AI backlash has a physical flashpoint now, and that's a problem

AI used to feel weightless. A chatbot lived in a browser tab. A model announcement was a bunch of benchmark numbers. It all felt very abstract, and opt-in. If you didn't like AI, cool; you didn't use it, or at most, you talked smack on anyone who did. The cost of the whole thing stayed mostly invisible, behind the balance sheet of big tech or venture capitalists. This spending eventually ballooned to a point of concern for major institutional investors, and many ink-shaped pixels were spilled over whether or not we are in an AI bubble. But beyond the Monday morning quarterbacking on behalf of the global economy, there was still not really a physical manifestation of AI for its haters to rally behind.

Then the servers arrived.

Data centers have turned AI into local politics (and as the saying goes, all politics is local). Now that they are expanding as rapidly as they are, they need land, electricity, transmission lines, water, cooling systems, backup power, tax incentives, zoning approvals, and local city council consent to continue growing. A lot of times, this is despite the concern from residents who did not ask for a hyperscale computing campus down the road. 

In fact, Gallup just found that seven in 10 Americans oppose AI data centers in their local area, including 48% who strongly oppose them. In the same survey, data centers were more locally unpopular than nuclear plants.

This opposition is broad. Gallup reported majorities across party groups, regions, ages, races, education levels, incomes, and urbanicity opposing local AI data centers. The objections are also concrete: water use, energy use, pollution, traffic, quality of life, higher utility bills, and taxpayer costs.

Data Center Watch, a research project tracking local opposition, says opposition to data centers consolidated into a national political force during the second half of 2025. Its March 2026 updates included a Pennsylvania borough denying a data center campus in a 5-0 vote, a North Carolina town moving toward a one-year moratorium, and South Dakota signing new infrastructure-cost requirements for large data centers.

The backlash has gotten so popular, Erin Brockovich (from that Julia Roberts movie fame) has a new AI data center map that is tracking the growth of data-centers as it unfolds town by town, with some projects welcomed and others delayed, contested, or abandoned. The site lists 33 operational AI data centers, 44 under construction, and 27 proposed as of its May 24 update, while a Nieman Lab writeup said the project had already collected 2,716 community reports of incoming datacenters. In a way, you can look at this as a way to track where the backlash will spread next.

If you've never heard the phrase NIMBY before, it stands for "Not in My Backyard", and that's definitely what's going on here; even reasonably pro-AI people can turn anti-datacenter when it finally shows up in their "backyard."

This is the part of AI politics Silicon Valley often struggles to process. People may like tools that make work easier. But then they may still oppose the infrastructure that makes those tools possible when the physical manifestation (including the costs) show up on their street (or grid, or local water system, or local tax bill).

The White House sees the same infrastructure boom, but through a different lens. In July 2025, it issued an executive order to accelerate federal permitting for AI data center infrastructure, defining covered data center projects as facilities needing more than 100 megawatts of new load for AI inference, training, simulation, or synthetic data generation. This means the federal government wants faster buildout for the enormous server farms that run modern AI.

Naturally, those two divergent perspectives has created the current collision course. The industry needs infrastructure faster. Many communities want a veto, or at least a much louder voice. And the State now has to decide how to respond to this public outrage while at the same time balancing the delicate equilibrium of the AI race.

It might be helpful to think of the current AI race as the "space race" of the 1960's to 1980's, but of the 2020's through the 2040's. The space race was unquestionably a geopolitical and psychological win for the United States. CFR notes that Sputnik convinced U.S. policymakers the country had fallen dangerously behind, pushed Washington to invest in education and scientific research, and helped plant seeds for future innovation and competitiveness. You can think of DeepSeek and "the DeepSeek moment" as AI's sputnik. We'll come back to this idea later on in the story.

The State is starting with security

WIRED's investigation shows one version of the state response: surveillance, threat assessment, and new domestic extremism categories aimed at following "anti-tech" violent extremism.

The article says WIRED obtained more than 1,000 pages of unpublished reports from the Department of Homeland Security, FBI, and fusion centers showing a shift toward monitoring “anti-technology extremists.” The reports appear in the wake of attacks on CEOs, data center protest movements, and rising fears about AI job replacement.

One New York Intelligence and Counterterrorism Bureau report warned that AI's disruption over the next five years could fuel large-scale protests that devolve into civil unrest and “anti-tech violent extremist activity,” especially in large cities. WIRED notes that the term “anti-tech violent extremism” does not appear in publicly available DHS or FBI domestic extremism reports or guides.

Categories have power. Once a label exists inside intelligence systems, it can shape what police look for, what private intelligence firms monitor, what activists fear, and what ordinary protest gets associated with.

Some of the concern is grounded. WIRED points to extremist strains that have plotted or carried out violence, including insurrectionary anarchists, eco-extremists, and ecofascists described by extremism researcher Mauro Lubrano. The article also cites SITE Intelligence, a private open-source intelligence firm, saying it has seen a spike in online threats advocating sabotage against data centers.

The credible counterargument is simple: threats to people and critical infrastructure deserve attention. Data centers are now part of the economy's nervous system. AI models are also becoming powerful enough that government officials worry about cybersecurity risks.

The Trump administration even drafted an AI executive order that would have created a voluntary review system where companies could give agencies an early look at frontier AI systems up to 90 days before release, according to The Washington Post, who also reported that Trump paused the order after Silicon Valley leaders warned it could slow U.S. AI development. Think of the space race here as well.

The civil-liberties concern is just as simple: broad categories stretch. WIRED reports that fusion center materials included constitutionally protected demonstrations and local civic meetings, and that suspicious-activity indicators can include ordinary behaviors like photography, observation, and testing security. The article also says a video from the progressive nonprofit More Perfect Union about a data center's local effects circulated in fusion-center materials even though the video did not advocate violence.

That is the risky path: a real fringe produces a broad surveillance frame, and the frame begins to cover normal dissent. It could also have the complete opposite effect: more distrust of big tech and AI datacenters because of the broad surveillance they are, in this scenario, enabling.

The Church is drawing a line in the silicon

Pope Leo XIV's response begins somewhere else entirely.

Magnifica Humanitas opens with two biblical building projects. Babel is the tower built for power, uniformity, and self-assertion. Jerusalem is the city rebuilt through shared responsibility, prayer, work, listening, and the contribution of many hands.

The Pope's AI argument follows from that image. Technology can heal, connect, educate, and protect. It can also divide, exclude, and create new forms of injustice. The primary choice is therefore the kind of city the technology is building.

His Babel is basically the ultimate techno-corporate state: a system where data, compute, platforms, patents, algorithms, and infrastructure are concentrated in the hands of a few; where efficiency becomes the measure of human worth; where AI decisions appear neutral while hiding the values of their creators; where communities receive finished systems rather than participating in the choices that shape them.

His Jerusalem is slower and less shiny. It is a world where technology remains answerable to human dignity, local participation, social justice, work, education, family, truth, peace, and the common good. It is less demo day, more city council meeting. Which, admittedly, is a rough pitch deck. Unless you're my parents, when was the last time YOU went to a city council meeting (jk, I'm sure many of you reading this do go to your city council meetings, and mad respect to those of you who do, because sadly, I do not go to mine, so good for you!).

The encyclical's most useful sentence for AI policy is this: AI cannot be treated as morally neutral. Technical systems embody choices through what they measure, ignore, optimize, and classify. If a system excludes people without appeal, the moral problem has already entered the design.

That point connects directly to current AI fights:

• Hiring systems decide who gets seen.
• Credit systems decide who gets access.
• School systems decide what gets learned.
• Social platforms decide what speech gets amplified.
• Data centers decide which communities carry the infrastructure costs for machine intelligence.
• And Frontier models decide which companies and countries own the next layer of power.

The Pope's answer is a demand for participation, accountability, and limits. He argues that responsibility must be defined at every stage, from developers to deployers to institutions that rely on AI. He calls for transparency, independent checks, equitable data access, routes for appeal, and public oversight.

FYI, we include all the key claims from the piece at the bottom of this article.

He also expands one old Catholic principle into a very modern policy claim: the universal destination of goods. In the encyclical, that principle now covers patents, algorithms, digital platforms, technological infrastructure, and data. That is a big move. It says the key assets of the AI age carry social obligations because they shape everyone else's life.

Jerusalem demands ownership, not just participation

The phrase “participation” can sound like a town hall meeting where everyone gets three minutes at a microphone before the decision that was already made. That model of public appeasement will fail.

Participation has to include upside for the participants (ideally, all participants).

This is currently the missing piece leading to the data center backlash. Love it or hate it, AI infrastructure is one of America's key growth stories atm (I would argue it is the most important one). Goldman Sachs projects annual AI infrastructure spending could reach $800B by the end of 2026, with AI expected to contribute materially to capital expenditure growth. Investopedia reported Morgan Stanley's estimate that AI infrastructure investment will account for roughly one-quarter (!!!) of GDP growth this year.

That growth shows up in places people actually care about: jobs, tax bases, company earnings, markets, and even retirement accounts. The connection is uneven, but it is real. A Washington Post report said around 62% of Americans own stocks, and that AI-related stocks drove much of the market's 2025 gains. The same logic explains why BlackRock's Larry Fink warned that AI could widen inequality unless more people participate in the capital gains from AI; his argument was that prosperity feels distant when market capitalization rises, but ownership remains narrow; out of (portfolio) sight, out of mind.

Critics who treat data centers only as pure local extraction miss that AI infrastructure is part of the economic engine propping up U.S. growth and market confidence. Meanwhile, data-center supporters who treat opposition as just pure ignorance miss the obvious political problem: people oppose booms they experience as someone else's asset but their individual burden.

The practical answer to this issue is a new ownership bargain. Communities hosting AI infrastructure need ways to share in the upside (not just as tax revenue), see the benefits, and enforce the conditions. Some possible mechanisms include:

• Host-community benefit agreements: legally binding deals negotiated with community groups as primary counterparties that fund local schools, clinics, parks, air monitoring, apprenticeships, childcare, and emergency services.
• Utility-bill credits: automatic monthly credits for residents in host communities when data centers drive new grid investment or local power constraints.
• Local AI infrastructure dividends: a dedicated share of property taxes, lease payments, or power-sales revenue paid into a community trust with transparent rules.
• Community energy ownership: local residents own shares in solar, storage, geothermal, or grid assets built to serve the data center, similar to community-benefit energy models.
• Worker transition funds: data center approvals trigger local retraining dollars, apprenticeship seats, and hiring targets for workers exposed to AI displacement.
• Public AI wealth funds: states or the federal government place a slice of AI infrastructure taxes, spectrum-style auction fees, or excess profits into a public fund (like the sovereign wealth funds from oil revenues that many other countries have, for example) that pays dividends or invests in education.
• Windfall commitments: AI firms make ex ante commitments to share extreme AI profits for broad public benefit, an idea formalized in the Windfall Clause literature and in a way, made manifest in the public benefit corporate structures of Anthropic and OpenAI.

None of these are magic solutions, however. Community benefit agreements can become performative. Utility credits can be too small. Public funds can be raided. Windfall clauses can become moral branding without enforcement. Still, they all point in the right direction: AI legitimacy requires more than consent. It requires a stake.

This brings me back to the space race: Remember, America's perception of being ahead in AI matters for the economic picture just as much as the datacenter build-out. That's because it inspires investing in the US economy; of course you want to own US tech stocks because it's the best way to own AI.

For instance, the White House's AI Action Plan explicitly frames AI as a race for global dominance, built on three pillars: accelerating innovation, building AI infrastructure, and leading in international diplomacy and security. It also says the U.S. currently leads in data center construction, computing hardware performance, and models.

So how does this match up with the space race? In Moonshot: Public R&D and Growth, Shawn Kantor and Alexander Whalley found that NASA contractor R&D increased manufacturing value added, employment, and capital accumulation in space-related sectors. But they also found the iconic moonshot program had only modest economic effects for both local and national space-related sectors.

• A Phys.org summary of the research says NASA contracts produced large gains for receiving firms, often 35% to 50% in employment and capital investment.
• Those gains were highly localized, with little evidence of broad spillovers to neighboring industries or regions.
• The estimated fiscal multiplier was about 0.3, meaning each federal dollar spent increased local economic output by roughly $0.30, below the 0.6 to 0.8 range cited for general government spending.
• The safer conclusion is that the space race worked as mission-driven industrial policy and geopolitical signaling, rather than a broad economic miracle.

That is the warning for AI. Winning the AI race may matter enormously for national confidence, diplomatic leverage, military power, scientific leadership, market valuations, and the belief that America still owns the next industrial platform.

This time, the economic channel is already more visible than Apollo's (see the Goldman and Morgan Stanley projections shared above). That makes the perception of being ahead in AI economically important in a way the space race only partly was. It supports investor confidence, tech valuations, retirement accounts exposed to index funds, construction demand (important if you care about new housing builds too), energy investment, allied adoption of U.S. systems, and the broader story that U.S. growth still has a future-facing engine.

But of course, prestige does not automatically become shared prosperity. Apollo could win the moon and only produce modest broad multipliers. AI can win the leaderboard and still lose local dividends, especially if communities feel they are paying for the boom through higher bills, water stress, pollution, noise, and with no ownership stake.

The benefits also have to feel believable

Financial upside is only one part of the bargain. People also need a credible story of social upside, too.

This is where abundance rhetoric often misses the room. The claim that AI will create material abundance can sound exciting to people already positioned to capture it. It can also sound threatening to people who have watched previous booms arrive as rent increases, wage pressure, weakened unions, local disruption, and a more fragile civic life.

Some people doubt abundance is even possible because they have never experienced it. Some fear what comes with it: more automated decisions, more noise, more energy demand, fewer human jobs, more institutions telling them the future is inevitable, and less overall choice.

For the Pope's side, he says AI can help humanity if it remains answerable to human dignity. It can heal, educate, connect, reduce dangerous work, improve services, support environmental stewardship, and give people more room for care and creativity. In Anthropic terms, AI can augment human work rather than replace it, and public input can shape model behavior rather than leaving values to a few labs.

The political task is making that visible and, for lack of a better word, feel-able to people before resentment hardens. Communities need to see concrete benefits: lower bills, cleaner power, better schools, local training, transparent environmental data, real jobs (and not just new, but good ones; jobs people actually want to do), civic oversight, and tools that improve public services. Workers need paid pathways into AI-augmented work. Parents need credible child protections. Schools need resources to teach discernment. Local governments need technical help so they are not outmatched by hyperscaler lawyers.

That is the Jerusalem case for AI. The future has to look like a city people can live in, not a magic factory they are asked to stand outside of and keep plug quarters into if they want to keep getting the benefits from.

That buy-in needs to apply to the models themselves, too.

For the AI labs' part, Anthropic-backed researchers have explored Collective Constitutional AI, a process for using public input to shape model behavior. In that study, a model fine-tuned with collectively sourced principles showed lower bias across nine social dimensions while maintaining comparable performance on language, math, and helpful-harmless evaluations.

That is basically Jerusalem at model scale: the rules governing the system come from more stakeholders than just the builders.

Anthropic's own labor data complicates the fear story. A recent paper using Anthropic Economic Index data found that 78.7% of observed AI interactions were augmentation rather than automation. The jobs debate remains unsettled. Still, the data points toward the positive case: AI can help people do more complex work when institutions design around human agency instead of replacement.

The encyclical's positive case is broader. AI can support care, education, scientific discovery, environmental monitoring, safer work, better services, and more human time. The Pope's condition is that those benefits must be real for ordinary people. A future of abundance is hard to sell to someone who expects the abundance to arrive as a permanent upper-class, leaving them in the dreaded "permanent underclass."

The environmental case has to be answered too... we argue with engineering and power-sharing

Data center critics have a strong environmental argument because many harms are local, measurable, and already visible. Brockovich's reporting site organizes the complaints around six basic concerns: energy consumption, water usage, e-waste, location risk, scalability, and noise. That list turns abstract AI infrastructure into things a city council can inspect, permit, monitor, and charge for.

The clearest recent example of datacenter's negative impact on local communities is xAI's Colossus project in Memphis. AP reported that the NAACP and the Southern Environmental Law Center notified xAI of intent to sue over alleged Clean Air Act violations tied to gas turbines at the facility. Critics said the turbines could worsen pollution in nearby predominantly Black communities. The Guardian reported that EPA later found xAI had illegally operated portable methane gas turbines without proper air-quality permits. TIME reported that University of Tennessee researchers found a 79% increase in peak nitrogen dioxide levels nearby.

The broader public-health concern is bigger than one facility. A Business Insider investigation estimated U.S. data center pollution could impose $5.7B to $9.2B in annual public health costs, including up to 595 premature deaths and 300,000 asthma episodes. A research paper from December 2024 on AI's public-health burden projected U.S. data centers could create public health costs of more than $20B per year by 2030, with economically disadvantaged communities facing much higher household burdens.

So the answer to these issues has to be specific, enforceable, and local. “Trust us, you're going to benefit" is not enough and will also fail on its own.

Luckily, the fix list already exists:

• Replace temporary gas turbines with clean, firm power: solar and batteries help, but data centers also need geothermal, nuclear, long-duration storage, and clean power contracts that add new generation rather than reshuffle existing supply. The Trump administration is working to speed up new grid installations and power plant permitting; these type of efforts should be supported.
• Make data centers grid-interactive: a Phoenix field demonstration showed AI workloads could cut cluster power use by 25% for three hours during peak grid events while maintaining service quality.
• Use cleaner backup systems: batteries, fuel cells, more geothermal, stricter generator limits, and continuous emissions monitoring can reduce reliance on diesel or gas backup power.
• Require real permitting for temporary gas: temporary turbines should not become a loophole for long-term pollution. If gas is used, communities need permits, NOx controls, time limits, leak monitoring, and public emissions data.
• Site around carbon and water constraints: one 2026 study found AI compute demand is highly concentrated and can stress local power systems, while another found advanced cooling can reduce cooling energy by up to 50% and low-carbon, water-secure siting can cut combined footprints nearly in half.
• Reuse waste heat: hot-water cooling and district-heating models can turn server heat into a local asset for buildings, campuses, or industrial uses.
• Make health data public: air monitors, water reporting, generator-testing schedules, asthma and ozone impacts, and cumulative exposure studies should be conditions for approval, not favors after complaints begin.

The key is coupling technical fixes to community control. A cleaner data center that still raises local bills and hides its emissions will remain politically fragile. A data center that funds local power, publishes and works to mitigate its harms, shares its benefits with the people, and gives residents enforcement rights becomes a different civic object entirely.

In a way, they start to look more like public utilities... not that I'm saying that's the only answer. There's many problems with the current energy industry due to the way electricity is run as a utility. A third, hybrid path taking the best parts of both private and public control could be the best option if engineered correctly.

The real fight behind all of this is over legitimacy

The State and the Church are reacting to the same pressure point: people feel AI is being done to them.

Workers hear that automation will raise productivity, then watch job cuts arrive before retraining does. Towns hear that data centers are strategic infrastructure, then ask who pays for the power lines, the water, the noise, and the land use. Parents hear that AI will transform education, then watch their children encounter synthetic media, widespread cheating, addictive feeds, and persuasive chatbots faster than schools can adapt. Voters hear that AI is national destiny, then discover that many decisions are made by a handful of companies with proprietary models and private incentives.

That is why the Pope’s Babel / Jerusalem frame works so well here. According to the point of view of Magnifica Humanitas, Babel is the wrong kind of infrastructure project: one language, one technology, one direction, and one ambition, with the builders trying to “make a name” for themselves.

In the Pope’s reading, the problem is the spirit underneath the project. It is built from pride, uniformity, and self-sufficiency, so it sacrifices communion for scale. Everyone is technically aligned, but nobody is truly participating. The result is a giant achievement that collapses into confusion because it was never built around human dignity in the first place. Kiiiinda sounds like the perfect metaphor for the perils of chasing AGI without a plan for what to do with it!

Jerusalem is the opposite model. In Nehemiah’s story, the city is rebuilt by shared responsibility. Nehemiah prays first, studies the broken walls, listens to the people, assigns families their own sections, coordinates priests, artisans, leaders, and young people, and rebuilds relationships before rebuilding the stone.

That is the legitimacy project AI needs now. A data center imposed from above feels like Babel: a tower dropped into town by people with money, models, and permits. But a data center built with local participation, visible benefits, worker protections, environmental safeguards, and community ownership starts to look more like Jerusalem: still infrastructure, but infrastructure people can recognize as partly theirs.

The U.S. policy landscape is already wrestling with this legitimacy question as well. One White House order from December 2025 sought a minimally burdensome national AI framework and created an AI Litigation Task Force to challenge state AI laws viewed as inconsistent with federal policy. The same order instructed agencies to evaluate state AI laws and consider federal reporting and disclosure standards that could preempt conflicting state laws.

That is one path: reduce friction so the national AI buildout can proceed. Joe Public might not like it, but at least he knows the rules.

State and local politics are pushing another path: moratoriums, zoning fights, water and energy reporting requirements, infrastructure-cost rules, child-safety bills, algorithmic-discrimination rules, procurement limits, and model-safety frameworks. Some will be clumsy. Some will be captured by local fear. Some will be necessary because national politics moves too slowly. Millville, New Jersey recently voted to ban new data centers after residents raised concerns about health, infrastructure, water, utility costs, and community character. In Texas, even Republican officials have split over proposed data center pauses as grid and water concerns mount.

The question underneath all of it is bigger than compliance. Who gets a vote, and who owns a share?

Companies want speed. National-security officials and companies want protection. Communities want consent. Workers want stability. Parents want safeguards. The Church wants technology ordered toward human dignity. Law enforcement wants to separate dissent from violence before violence happens. The danger is that each institution will solve only the part of the problem... the problem it is built to see.

That is why we all must try to see with eyes wide open, to truly understand all the individual parts that make up the whole of the problem, and then try to take the next best action, knowing as much as we possibly can, to make the best action we possibly can given the facts on the ground and all the possible sides and scenarios and second and third order impacts of the decisions we make, be them economical, environmental, social, political, or anything in between.

What the Pope adds that lawmakers usually miss

Most AI policy debates treat the person as a rights-bearing user, consumer, worker, voter, or data subject. That is useful. It is also thinner than the Pope's view.

Magnifica Humanitas treats the person as embodied, relational, vulnerable, morally responsible, and impossible to reduce to output. AI can simulate empathy, friendship, care, and advice. It cannot suffer, forgive, love, bear responsibility, know the body, or understand from within what work and relationship mean.

That may sound theological until it becomes practical.

A system that ranks job applicants has to answer to people who can be harmed. A chatbot that comforts a lonely child has to answer to the difference between simulated care and real relationship. A platform that feeds outrage has to answer to what it does to public truth. A data center that raises local utility costs has to answer to neighbors who did not sign the compute contract.

The Pope also connects AI to labor in a way that many policy documents avoid. He argues that work is more than income. It is identity, responsibility, cooperation, family stability, and participation in society. AI adoption therefore needs social criteria: job protection, retraining, worker participation, and corporate measurements that include dignity of work.

As Sam Lessin put it, AI is a meaning crisis, and while the work we do provides income, it also provides meaning, and without work, we also lose part of our meaning.

That is the missing line in many AI roadmaps. A company can say a tool raises productivity, but a society still has to ask whether that productivity creates more human flourishing, or simply concentrates economic gains while leaving the rest of society to absorb the shock.

The security frame has a limit

The strongest version of the State response says this: anti-tech violence is real, data centers are critical, AI systems may create national-security risks, and public officials cannot wait for a catastrophe before tracking threats.

That argument deserves to be taken seriously.

The limit arrives when public frustration itself becomes suspicious. A town resident photographing a data center site belongs in a different category from someone planning sabotage. A nonprofit criticizing local infrastructure costs belongs in a different category from a violent extremist network. A worker protesting automation belongs in a different category from a threat to public safety. It should go without saying in the United States of America, but the first Amendment protects free speech; therefore, citizens of the US have the right to speak up when something wrong is being done. And actually, criticism is the best user feedback you can get as a product developer... so actually, as technologists and a government supporting technology, we want healthy criticism to make better, more optimized products that people actually want. You don't want to discourage this... use the negative sentiment as signal to improve your AI products and governance for the better.

That said, AI backlash will continue to grow because AI is touching more of life. Treating that backlash mainly as a security threat will likely harden it. Treating it as a legitimacy problem creates more options. Treating it as user feedback will help create better technology, and hopefully, a more thriving civilization to benefit from said technology.

But what is legitimacy, really? Legitimacy means people believe any process is fair enough to accept outcomes, even when they dislike them. That requires visible rules, local participation, honesty around costs, the right to appeal, a viable path for labor transitions, child protections, timely environmental reviews (not just burdensome, bureaucratic red tape like the kind that prevent renewable plants from being integrated), public benefit-sharing (giving everyone buy-in on the upside), and a credible dialogue around concerns that distinguishes true violent threats from useful democratic dissent.

This is where the Church and the State accidentally meet. The Pope's Jerusalem story is a legitimacy model. It says the city is rebuilt with people, through institutions, under moral limits. The State can protect infrastructure. The harder job is making the infrastructure worth protecting, and worth living near. Hard does not equal too difficult to do. It means worth doing well. And who else, if not the state, is built to do hard things?

ICYMI: The Magnifica Humanitas Arguments, In Brief

We wanted to include this summary of the key parts of the encyclical because we found it very illuminating, but of course you should read the whole thing if you're interested. It argues that AI is forcing humanity to choose between Babel and Jerusalem: a world built for domination, efficiency, and private technological power, or a world rebuilt through dignity, truth, work, solidarity, justice, peace, and shared responsibility.

The whole document in one pass

• The central metaphor: AI is the “construction site” of our era. Humanity can build Babel, a tower of pride, uniformity, domination, and dehumanization, or Jerusalem, a shared project of rebuilding where everyone has a role and God, dignity, and the common good remain central.
• Technology is neither evil nor neutral. It can heal, educate, connect, and protect creation. But every technology reflects the priorities of whoever designs, funds, regulates, and uses it.
• The real issue is power. AI is developing mainly through private, transnational actors with resources larger than many states. That creates a new kind of private technological sovereignty over data, platforms, visibility, labor, and access.
• Regulation matters, but regulation alone is too shallow. The Church says society also needs spiritual, cultural, ethical, and political discernment: who holds power, toward what end, and with what view of the human person?
• The document is built on Catholic Social Doctrine. It restates the core principles: human dignity, human rights, common good, universal destination of goods, subsidiarity, solidarity, social justice, and integral human development.
• Human dignity is not earned. A person’s worth does not depend on productivity, intelligence, ability, efficiency, wealth, health, usefulness, or achievement. This is the key anti-technocratic claim.
• Human rights matter because they express that dignity. The document treats the Universal Declaration of Human Rights as a major achievement, while warning that rights become fragile when society forgets their foundation.
• The common good is more than everyone pursuing their own interest. It is the social condition that lets people and communities flourish together, especially the vulnerable.
• The “universal destination of goods” now includes digital goods. The document explicitly expands this principle to patents, algorithms, digital platforms, technological infrastructure, and data. Concentrating them in a few hands creates a new injustice.
• Subsidiarity becomes a digital governance principle. Communities, schools, universities, civil society, churches, workers, and local groups should have real input into systems that affect their lives. Decisions about data, algorithms, jobs, access, and platforms should not be imposed opaquely from above.
• Solidarity now includes the digital ecosystem. Data, AI, platforms, and computational infrastructure should be managed with attention to future generations, excluded communities, hidden workers, and poorer countries.
• Social justice must be built into technology from the start. The document rejects “deploy first, fix harms later.” Justice should shape design, data, access, oversight, and accountability before systems enter public life.

What it says about AI specifically

• AI is not human intelligence. It imitates some functions of human intelligence, often at great speed, but it has no body, experience, moral conscience, love, responsibility, suffering, forgiveness, or inner growth.
• AI systems are “cultivated” more than fully “built.” Even developers do not fully understand their internal representations and computational processes. That creates a scientific and moral duty for deeper research and humility.
• AI can create the illusion of objectivity. Its outputs may seem neutral, but they reflect the assumptions, data, incentives, and blind spots of designers and trainers.
• AI can simulate empathy, friendship, and love, but it cannot create real relationship. The danger is especially serious where people lack human connection; users may lose the desire for genuine relationships.
• AI adoption creates environmental costs. The document flags energy use, water use, CO2 emissions, storage, data centers, machines, cables, and infrastructure as part of AI’s real-world footprint.
• AI in consequential decisions requires accountability. Employment, credit, public services, reputation, opportunity, and freedom should never be handed to opaque systems without explanation, appeal, oversight, and responsibility.
• The document rejects the idea that AI is morally neutral. Every system embodies choices: what it measures, ignores, optimizes, classifies, rewards, and excludes.
• “Alignment” is not enough. The Pope argues society cannot merely ask tech companies to align AI with “human values” if those values are defined by a few private actors. Ethical frameworks themselves need public debate and social justice.
• Data should not be treated as purely private property. Since data is generated by many people and communities, it should be managed creatively as a common or shared good.
• The document names “new monopolies of AI.” These are concentrations of data, compute, regulatory influence, and model-building capacity that create epistemic, economic, and political asymmetry.
• “Disarm AI” is a major phrase. It means freeing AI from the arms-race mentality of bigger models, larger datasets, monopoly control, geopolitical dominance, and commercial supremacy. Disarming AI does not mean rejecting it; it means preventing it from dominating humanity.
• Developers have a spiritual responsibility. Every design choice reflects a vision of humanity. Developers are called to transparency, responsibility toward affected communities, and serious attention to the good being cultivated.

The human vision underneath it

• The Pope rejects transhumanist and posthumanist dreams when they treat the human person as something to surpass. Technology can heal and help, but humanity should not be reduced to a defective machine waiting for upgrades.
• Limits are part of being human. Illness, aging, vulnerability, suffering, weakness, and failure are not merely bugs to eliminate. They are also places where compassion, wisdom, love, dependence, and relationship become visible.
• The “more than human” is not technological enhancement. In Christian terms, the authentic “more than human” is grace: becoming more fully human through love, communion, and relationship with God.
• A civilization is judged by its care, not its power. The measure is whether it recognizes people as faces, not functions.
• The key test for AI: Does it make human life more human, more dignified, more relational, more just, more peaceful, and more open to the common good?

Truth, education, and public life

• Truth is a common good. Democracy depends on shared commitment to facts, verification, responsible argument, and trust.
• AI supercharges disinformation. Deepfakes, synthetic media, biased narratives, manipulated images, and algorithmic amplification make truth harder to protect.
• Digital platforms shape culture, not just information. Whoever controls the feeds shapes the collective imagination: what people desire, fear, normalize, and consider real.
• The document calls for an “ecology of communication.” That means transparency in content-ranking systems, protection of personal data, stronger journalism, better public forums, and education in verification.
• The Church says it must model transparency too. It explicitly praises journalists who exposed abuse and says the Church must not wait for others to force uncomfortable truths into the open.
• Education should teach when not to use AI. The document says AI literacy is not just learning how to use tools, but knowing when restraint, slowness, thought, and human effort matter more.
• Schools should not follow the pace of the digital world. Their job is to provide what digital systems cannot: shared time, trustworthy relationships, reflection, silence, deep study, moral formation, and love of truth.
• Children need stronger protection online. The document warns about early unsupervised device use, addiction, grooming, blackmail, cyberbullying, sexual exploitation, violent content, pornographic content, fake profiles, and AI image/video manipulation.
• Parents cannot solve this alone. Lawmakers, schools, families, and platforms need shared responsibility, including age limits, platform accountability, and protections against online exploitation.

Work and the economy

• Work is central to human dignity. It is more than income; it forms identity, relationships, responsibility, creativity, family stability, and participation in society.
• AI should relieve dangerous, repetitive, or degrading work, but not make humans disposable. The document warns that AI can de-skill workers, intensify surveillance, force humans to match machine rhythms, and reduce agency.
• Unemployment is a grave social evil. The Pope treats job loss from automation as a threat to families, young people, local economies, citizenship, and social peace.
• The protection of employment should remain the rule. Profit cannot justify systematically sacrificing jobs because the person is an end, not a means.
• The document calls for “social criteria for innovation.” Every major introduction of automation or AI should include verifiable measures for job protection, retraining, and worker participation.
• Continuous training should be accessible to everyone. The cost of adaptation should not fall only on individuals.
• Companies should measure dignity of work as part of success. Productivity and profit are incomplete metrics.
• GDP is not enough. The document calls for complementary metrics that measure work dignity, shared prosperity, inequality reduction, environmental protection, and human wellbeing.
• Finance must serve real work and development. It criticizes finance for its own sake, financial speculation detached from moral foundations, and capital income replacing labor income.
• The “invisible hand” is insufficient in the age of AI and robotics. Politics must guide markets and technologies toward dignified work, inclusion, and equitable distribution of innovation’s gains.
• Concrete economic criteria: algorithmic decisions must be transparent and contestable; innovation must include access and skills investment; taxes, social protections, and industrial policy must correct concentration of wealth and power.

Families, youth, freedom, and addiction

• The family is treated as a primary social good. Economic and technological systems that destabilize work also destabilize family life.
• Young people suffer uniquely from job insecurity. Work helps them form identity, relationships, responsibility, and vocation. A world of unstable work blocks human development.
• Digital addiction is a moral and social problem. Platforms monetize attention and weakness, often treating users as means rather than ends.
• Freedom is now a public digital issue. If data systems profile, predict, rank, and influence people without transparency or recourse, freedom is undermined.
• Control happens through visibility. What platforms amplify, hide, reward, or punish shapes behavior, opinions, conformity, and self-censorship.

Slavery, exploitation, and digital colonialism

• The document directly links AI to hidden labor. Data labeling, model training, and content moderation depend on millions of mostly invisible workers, often under harsh conditions and low wages.
• AI also depends on extractive supply chains. Devices and chips require rare earths and raw materials, sometimes produced through dangerous labor, including child labor.
• The digital economy can enable trafficking. Platforms, messaging systems, anonymous payments, and profiling tools can help recruit, control, and move victims.
• The Pope asks pardon for the Church’s historical delay in fully condemning slavery. He treats that memory as a warning: society must not later need to apologize for ignoring today’s digital and labor exploitation.
• Digital colonialism is a key idea. The document says colonialism now also appropriates data: health data, genetic maps, demographic information, epidemiological profiles, and other datasets collected from vulnerable regions.
• Health and population data are called the new “rare earths” of power. Whoever controls them can shape markets, investments, medicines, protections, and who counts.
• Action items on slavery/exploitation: make digital supply chains transparent; require ethical due diligence; protect workers; assess social impact of data-driven business models; make platforms cooperate against trafficking and exploitation.

War, AI, and peace

• The final chapter argues that AI makes war faster, more impersonal, and easier to normalize.
• War is being culturally normalized again. Rearmament, polarizing media, regional conflicts, historical amnesia, and algorithmic outrage are making war seem more acceptable.
• The document says “just war” theory is now outdated in practice. It argues the theory has too often been stretched to justify wars, preventive attacks, and violence with grave civilian costs.
• AI weapons are a non-negotiable concern. The Pope rejects delegating lethal or irreversible decisions to opaque or automated systems.
• No algorithm can make war morally acceptable. AI can only accelerate and depersonalize conflict if used to optimize violence.
• AI warfare criteria: responsibility must be traceable; decision chains must be reconstructable; lethal force must remain under effective human control; speed must not dominate moral judgment; civilians must be clearly protected.
• The document calls for international rules on AI weapons and cyber conflict. Cyberattacks, data manipulation, influence campaigns, and AI-enabled destabilization require diplomacy, attribution norms, and civilian protections.
• The culture of power is the enemy. It includes arms markets, military-industrial incentives, nuclear deterrence, hybrid warfare, “might makes right,” enemy-based identity, and political realism that treats war as inevitable.
• The alternative is the “civilization of love.” This means translating charity into institutions, justice, diplomacy, solidarity, development, disarmament, and care for victims.

The five practical paths it gives

• Disarm words. Reject verbal aggression, dehumanization, propaganda, hatred, ridicule, and the “war of words and images.”
• Build peace through justice. Peace is not merely the absence of conflict; it comes from just relationships, institutions, and treatment of the vulnerable.
• Adopt the perspective of victims. Do not analyze war only from strategy rooms. Look at civilians, children, hospitals, schools, refugees, and the wounded.
• Cultivate healthy realism. Avoid both fantasy idealism and cynical resignation. Realism should identify constraints and interests in order to make peace possible.
• Revive dialogue and multilateralism. Diplomacy, negotiation, interreligious dialogue, the UN, and reformed international institutions are necessary to prevent conflict and rebuild trust.

The conclusion’s program for Christian life

• Stay faithful to truth. Resist manipulation, algorithmic persuasion, and the reduction of reality to usable material.
• Invest in education. Teach people, especially children and young people, to use digital tools responsibly and freely.
• Cultivate relationships. Protect meals, community, presence, service, tenderness, and care from being replaced by mere connection.
• Love justice and peace. Examine AI supply chains, hidden labor, war profiteering, manipulation, and concentration of power.
• Enter the construction site. The Pope calls Christians and people of goodwill into labs, companies, schools, media, institutions, and local communities to rebuild what is collapsing.

The Pope’s core argument is not “AI bad.” It’s “AI is infrastructure now, and infrastructure always reveals what a society worships.” If the goal is efficiency, profit, domination, and control, AI becomes Babel. If the goal is dignity, truth, work, justice, care, and peace, AI can become part of rebuilding Jerusalem.

• His sharpest tech-policy claim: data, algorithms, platforms, compute, and AI infrastructure should be treated partly as common goods, not merely private assets.
• His sharpest labor claim: AI adoption should be judged by whether it protects work, retrains people, shares gains, and strengthens families, not just whether it raises productivity.
• His sharpest moral claim: a machine can simulate intelligence, empathy, and care, but it cannot bear responsibility, love, suffer, forgive, or recognize good and evil.
• His sharpest governance claim: society needs the power to slow AI down when acceleration strips communities of participation, accountability, and recourse.
• His sharpest peace claim: autonomous or opaque systems must never make lethal decisions, because speed and optimization cannot replace conscience.

The final takeaway: AI is testing whether humanity still believes the human person is sacred before productive, relational before computational, and responsible before powerful.