From Digital Gods to Data Centers: AI, Infrastructure Ethics and the Accountability Gap

AI-Generated Religion: Digital Avatars, Robotic Priests and the Outsourcing of the Sacred

A few days ago, I ended up being pulled into this rabbit hole and the blame goes to my grandma. My grandma is 94 and has fallen for phone scams more than once. She's also deeply religious, so when I came across a certain post my first thought wasn't about tech, but about her. About who that's built for, and what it actually does to people who genuinely can't tell the difference, or maybe don't even want to.

There are some pretty big and also pretty quiet shifts between AI and religion, and I don´t know if I´m the only one, but I had no idea they were happening. Spiritual practice, ritual authority, religious experience: all of it is increasingly being filtered through screens, algos and fake voices. Various methods that are already being used by real people across commercial platforms, religious institutions and experimental settings worldwide.

One of the clearest examples and also the first I´ve been confronted with about this whole topic is the rise of AI-generated religious avatars built for personal interaction. There is a US-based company called " Just Like Me " offering video- call access to a synthetic Jesus avatar for around $1.99 per minute. What makes this particularly striking is not the price point, but rather the emotional response. The CEO has described how users often develop genuine emotional bonds with the system. These interactions are not perceived as symbolic or playful; for many users they feel real. That raises tons of uncomfy questions about whether religious experience can be partially simulated and what it actually means to form emotional attachments to something that cannot reciprocate.

The same shift is also happening in other institutional and ritual contexts. In Japan, an AI system known as "Emi Jido" , also being called a non-human Buddhist priest, has reportedly been ordained by a Zen practitioner. This appears to be the first time a religious legitimacy is being formally extended to an artificial agent. A separate project called "Buddharoid", developed through collaboration between academic institutions and technology ventures, takes this further by attempting to physically embody ritual authority in a humanoid robot.

Europe isn't far behind either. A Lutheran church in Finland ran a full christean AI-assisted service; in other words, algorithmically written sermons, AI-generated music and an avatar leading the whole congregation. There has also been a rabbi chatbot being trialed to answer theological questions in real time, turning centuries of religious interpretation into something like ChatGPT.

All of these developments have triggered a growing ethical and theological debate and many institutions have a big opinion on it. Pope Leo XIV has described AI as one of the most critical issues facing humanity, warning that it may negatively affect intellectual, neurological and spiritual development. That's not even a sloppy take anymore. It reflects a much broader concern that AI mediated interaction may end up altering not just religious practice but also our deeper cognitive and formative processes through which belief systems are built in the first place.

From a Buddhist philosophical perspective, some researchers have raised similar concerns around the centrality of effort in spiritual cultivation. Practices like meditation, discipline and sustained ethical training aren't just helpful tools you can swap out for something more convenient. They are constitutive of the development itself. The argument is that AI systems designed to reduce cognitive or emotional effort may quietly undermine the transformative nature of religious practice, not by replacing it outright, but by easily making it in a way that fundamentally changes what it actually is.

That said, I'm not here to pass judgment on emerging belief systems or new forms of religious practice. Spirituality has always evolved alongside technological and cultural change and it would be way too easy to frame all of this as either authentically human or hopelessly distorted. The reality is probably messier and more interesting than either of those positions and honestly that's exactly what makes it worth paying attention to.

At the same time, it's worth asking a harder question: do any of these experiments actually make sense within the religions they claim to represent?

Because there's a huge difference between technology being used to support a belief system and technology quietly redefining what that belief system even is. When an AI puts on the face of a sacred figure or spiritual guidance becomes something you access like a streaming service, the question isn't just whether it's technically impressive, but rather whether it's theologically coherent at all. Whether the religion would even recognize itself in the mirror these days.

Data Centers and Moral Responsibility: Built to Scale, Not to Be Responsible?

Large tech companies like Microsoft and Meta are now pouring more money into energy infrastructure than the whole entire US electricity sector combined. That's usually framed as an economic story - but look a little closer and it starts to feel like something else. Vast amounts of capital and energy, continuously fed into systems whose outputs are becoming increasingly abstract: language models, predictive tools, synthetic media, autonomous decision- making. At some point the scale of it stops making sense as a purely economic decision, though I'm not sure what the right word for it is either.

The language our industry uses doesn't exactly push back on that reading. Terms like "AGI," "Singularity," and "Alignment" carry a pretty big weight that goes well beyond the whole tech jargon. They echo older structures like salvation, finality and moral order. With that mindset, AI isn't just being built, but rather anticipated and framed as the thing that will eventually resolve complexity, uncertainty and maybe even the messiness of being human. Taking all facts into consideration, that's closer to theology than the actual engineering language.

This quasi-religious structure has a physical dimension too. DCs are often built in deserts, arctic regions, or remote industrial zones, places deliberately far from where people actually live. Like monasteries, they concentrate energy, discipline and continuity in isolated environments. If I´d unleash my poet, I´d say both systems withdraw from society while continuing to shape it from a distance. That contradiction gets way more intense when you look at something like the AI generated religious avatars I mentioned earlier. The interaction feels intimate, even spiritual, and yet behind it sits infrastructure that consumes energy on the scale of entire countries. The user experiences emotional immediacy while the material reality stays completely invisible and that gap is not accidental. Tech companies tend to resolve this tension through sustainability narratives. Carbon neutrality pledges, clean energy commitments, green transition roadmaps all suggest that a kind of ecological redemption is already underway. But the same infrastructure frequently still depends on fossil fuel sources, quietly sustaining the emissions it publicly claims to be moving beyond. Result is a clear structural contradiction between symbolic decarbonization layered over material dependency.

At the core of this whole debate sits one powerful assumption: that data centers are neutral infrastructure. They get compared to electricity grids or water utilities, pure enablers of whatever happens downstream. But from what I´ve seen in this industry, is that this analogy is starting to fall more and more apart. Infrastructure neutrality isn't some natural property of the technology, it's also a political construction and historically the same argument has been made in telecoms, logistics and defense, where the line between enabling and enabling responsibility has always been contested. As computational demand pushes into autonomous weapons systems, mass surveillance architectures, synthetic propaganda networks and biological simulation models, DCs stop being passive infrastructure and become active preconditions of capability. They don't just host computation, they also define what kinds of computation are possible at scale which is a whole different thing. Some limits already exist, but they're driven by legal exposure and reputational risk rather than any coherent ethical framework.

When AWS pulled the plug on Parler after the Capitol events, or Microsoft restricted services to Russian state media after the invasion of Ukraine, those decisions looked like moral interventions. In reality they were risk management calls made under conditions of high political visibility. The moral framing was a byproduct, but not the motivation behind it (which doesn´t mean the outcome was wrong). Meanwhile, vast areas stay structurally under-regulated. Compute allocation for autonomous weapons research, state surveillance systems, industrial-scale disinformation, in these domains the absence of action quietly functions as permission. Nobody said yes, but nobody said no either. Which brings us to the central governance problem. If DCs were to enforce moral limits, who gets to define them? In practice that authority would concentrate in a handful of US-based corporations, Microsoft, Amazon, Google, Meta, Oracle, giving a de facto oligopoly veto power over global technological possibility. That's not even close to a democratic ethics framework. The alternative, state regulation, comes with its own risks. In authoritarian contexts, infrastructure control becomes a tool of censorship and surveillance fast, but obviously not as an existing clean governance model, but rather as different configurations. At least that´s my take on it so far.

Small Data Centers as Moral Actors?

A small data center operates less like Amazon and more like a landlord. The operator usually knows their clients personally, stays in direct contact and in many cases has at least a rough idea of what their infrastructure is actually being used for. That creates a very different moral environment than what you find at hyperscale providers, where millions of anonymous workloads run across global infrastructure and nobody is really looking.

At the smaller scale, ethical questions can get uncomfortably concrete. What do you do when you know a client is running stalkerware on your servers? Or deploying an AI system built to generate targeted disinformation about local politicians? Or processing datasets for autonomous weapons development? These aren't hypotheticals being debated in some corporate ethics committee. They're immediate, situational and sitting right in front of you.

The problem is that proximity doesn't automatically translate into action. Small DCs operate in brutally competitive markets where pricing pressure is often decisive, and turning away a client doesn't make you ethical, it just makes you the provider that asked too many questions while your competitor quietly took the contract. It's the same race to the bottom logic that shows up in banking, chemical disposal and tax advisory; industries where ethical restraint gets structurally penalized by market competition.

Without industry-wide standards or legal frameworks to level the playing field, acting morally becomes economically irrational and that's really the central paradox. Moral responsibility is most personally felt at the small scale, where the operator knows exactly what's running on their hardware, yet least capable of producing any meaningful systemic change.

Flip to the hyperscaler level and the problem runs in the exact opposite direction. The impact is enormous but moral proximity has dissolved almost entirely into abstraction, committees and organizational distance. Nobody is personally on the hook for anything and the incentive structures don't exactly encourage them to be.

The result is a twofold failure: Individual virtue without systemic coordination produces negligible change. Concentrated systemic power without democratic accountability produces significant risk. The small provider can act ethically but lacks leverage while the hyperscaler has all the leverage but operates under incentive structures where ethical judgment is rarely the deciding factor. Neither level resolves the problem. Together they just expose how thoroughly morality, market incentives and tech infrastructure have become misaligned in certain aspects.

Small DCs vs. Hyperscalers: The Ethics Gap

There are some pretty big ethical differences between a normal data center and a hyperscaler apart from scale. They are pretty much about a completely different relationship to responsibility, so I've broken them into a few aspects to make sense of it for myself.

Proximity to moral responsibility

Small data centers operate up close. Operators often know their clients personally and have at least some sense of what their servers are actually being used for. That creates a kind of immediate moral exposure where responsibility isn't some abstract policy question, it's sitting right in front of the provider.

Hyperscalers work in the opposite condition. Millions of anonymous workloads, no real human proximity and a big knowledge barrier that is partly justified through data protection rules but also serves a different purpose: keeping a comfortable distance. What looks like ignorance is often a deliberate design choice, one that protects legal and reputational boundaries a lot more reliably than it protects any actual moral clarity.

The outcome is a pretty big split. The people with the least power to change anything feel the most personally implicated. The ones with enough leverage to actually matter have built their entire operating model around not having to.

The result is very paradox: moral awareness is higher where impact is smaller, and lower where impact is maximal.

Can the rule apply to everyone?

From a Kantian perspective, the question is whether a rule of unrestricted hosting could actually be universalized. At the small scale, the answer is pretty clearly no. If every operator refused to impose any limits, you'd get a race to the bottom almost immediately. But at least that failure is visible and personal. The operator sits with the contradiction between principle and survival in a very direct way.

At the hyperscale level, the same logic gets distributed across global infrastructure until no single person experiences any moral friction at all. Ethical failure becomes diffuse rather than personal, which makes it less visible but not less real.

Who actually has the leverage to perform?

From a utilitarian perspective, the picture partially reverses. A small data center turning away a harmful client doesn't actually change that much, because that client will just go to the next provider down the street. The net benefit is close to zero and the ethical restriction becomes more or less pointless at a systemic level.

Hyperscalers are a different story. When AWS shuts down a platform or Google cuts access to computational resources, it can actually stop harmful activity at scale. The leverage is real. But these interventions tend to happen inconsistently and are triggered by reputational pressure rather than any principled ethical framework, which means the potential is there but rarely fully used.

In other words, impact is highest where ethical consistency is weakest.

Who´s actually responsible?

At a small scale, ethics is pretty personal. There's no compliance team, no buffer - just you, your team and decisions right in front of you. The way you act reflects directly on who you are as an operator like unfiltered accountability.

At the hyperscaler level, that changes completely. Moral decisions get distributed across committees, legal departments and compliance frameworks until no single person is really responsible anymore. Doing the right thing becomes more of a procedure rather than a conviction. Ethics often gets reframed as risk management, integrity becomes reputation strategy which doesn´t result in the absence of morality, but rather organizational depersonalization.

Structural power and control

With a smaller DC, ethics becomes a luxury that´s not always affordable. Mostly, there is little leverage, clients can leave at any time and doing the right thing often just costs you money.

However, the big players are a different story. Companies like AWS, Google, or Microsoft basically run the internet at this point. They have enormous power to shape what happens online, but they answer to shareholders, not society. The result is an interesting tension: all the influence, but without a clear framework for accountability. And since there's no real financial incentive to intervene ethically, that power mostly stays on the sidelines.

This produces power without morality and morality without power.

Is there even a solution?

Effective ethical governance at the small scale would require some form of external coordination, industry standards, regulatory frameworks, something that levels the playing field so that acting responsibly doesn't just mean losing business to the competitor next door.

At the hyperscale level, self-regulation doesn't really work because the incentives point in the wrong direction and state regulation brings its own risks, particularly in political contexts where infrastructure control can quickly become a tool for something else entirely. Both solutions solve one problem and create another. There´s no clean answer, just different trade-offs depending on what you're most willing to live with.

The recap of it all: a dual collapse of moral structure

Across both scales, ethics fail, but in pretty opposite ways.

Small data centers present high moral awareness but low systemic impact - Hyperscalers present high systemic impact but low moral proximity.

One can't do much even when it wants to. The other could, but the incentives don't really push in that direction, at least not consistently. Together, they don't point to a lack of morality, but to its conditions being spread too thin for anyone to hold across technological scale. Not saying that people are malicious, it´s just our system that isn´t built for that kind of accountability.

In this system, infrastructure does not merely distribute computation. It redistributes moral agency itself, splitting up responsibility until it´s either too small to matter or too large to feel.

And in parallel, a cultural layer emerges in which AI systems begin to occupy roles traditionally associated with religion: avatars of sacred figures, algorithmic rituals and computational forms of spiritual interaction. Whether this represents adaptation or transformation of belief systems remains an open question. What is already clear, however, is that both technological infrastructure and emerging AI-mediated spirituality are converging on the same structural condition: systems of immense power without a clearly locatable center of responsibility.

Data Protection vs. Moral Oversight of DC Operations: A Conflict Built Into the System

Now at the heart of this whole debate: moral responsibility requires knowledge, and data protection law restricts access to exactly that knowledge. This is not a minor technical conflict, but rather a foundational contradiction built right into the system.

In Europe, this tension is basically codified. Under GDPR and telecommunications confidentiality rules, data centers typically operate as processors, meaning they handle data on behalf of clients, not for their own purposes. And that distinction is a big deal, because while operators might have technical access to systems and data flows, they are generally not allowed to actually look at what's in them without a specific legal basis. A hosting provider that starts systematically monitoring customer traffic to run their own version of a moral check would most likely be breaking multiple laws at once, GDPR provisions, professional confidentiality requirements, sector specific secrecy rules included.

Obviously, there are narrow exceptions. Concrete suspicion of illegal activity, lawful government requests, tightly scoped automated abuse detection. But none of that adds up to anything close to a general license for ethical oversight.

For small data centers this makes everything messier. Smaller operators tend to know their clients, have context and often feel a real sense of responsibility for what's running on their hardware. But that moral proximity gets structurally undermined by legal frameworks that prevent them from acquiring the knowledge they'd actually need to act on it. So you end up in this strange position where you're close enough to feel implicated, but legally blocked from doing much about it. Two normative frameworks, both completely legitimate, pointing in opposite directions at the same time.

For hyperscale providers, the same legal framework starts to work very differently. The claim that "we are not allowed to know what our customers are doing" bundles together technical architecture, legal obligation and strategic positioning into one very convenient narrative. And while it's true that providers don't have unrestricted insight into customer activity, they often have plenty of general contextual awareness, from public information, business relationships,or usage patterns. The line between "not allowed to know" and "not wanting to know" is not always clear and in practice it's probably not always meant to be.

That ambiguity points to something deeper. The core problem isn't really whether data centers should impose moral limits, it's that the system itself is designed in a way that diffuses responsibility before the question even gets asked. Data protection law does exactly what it's supposed to do, it protects user privacy, and in doing so it also makes meaningful oversight over harmful uses of infrastructure structurally difficult. Privacy and accountability end up positioned against each other in a way that prevents either from being fully realized. That's not a bug, it's a design choice and it's worth being honest about that.

Which brings us to the part nobody really wants to sit with. We live in a world full of high runners and prophets of the saying "don't hate the player, hate the game." But if we hate the game and really mean that, then it's on us to change the rules. That is what a functioning society is supposed to do, isn´t it?

---