The Vatican’s AI Statement and the Tower of Babel Problem for Enterprise AI

When Pope Leo XIV published Magnifica Humanitas, the Vatican’s statement on safeguarding the human person in the time of artificial intelligence, the central image was not a robot, a server, or a model benchmark. It was a construction site.

The encyclical frames the AI era through the choice between building a new Tower of Babel and building a city where human dignity, justice, and shared responsibility can endure. That is a powerful image for enterprise AI, because the biggest risk facing organizations is not only that AI systems may produce incorrect outputs. It is that AI can become so fast, so embedded, and so influential that people inside the organization lose the ability to understand how decisions are being shaped.

For business and technology leaders, the Vatican’s AI statement is useful because it gives language to a problem that is already appearing inside companies, governments, and regulated industries. AI is not just another productivity tool when it starts influencing hiring, credit, compliance, public services, customer treatment, legal analysis, engineering decisions, or access to institutional knowledge. At that point, AI becomes part of how an organization thinks, acts, and explains itself.

The enterprise version of Babel is not a dramatic technical collapse. It is a slower loss of shared understanding, where outputs keep moving through the system but accountability becomes harder to locate.

Why Babel is the right metaphor for enterprise AI

The Tower of Babel is often read as a warning about human arrogance, but it is also a story about coordination. The builders share a language, a project, and a direction, yet the project becomes detached from humility and responsibility. The problem is not construction itself. The problem is construction without the human maturity needed to govern what is being built.

That distinction matters for AI.

Most enterprise AI conversations still focus on capability: which model is more accurate, which assistant is faster, which automation can remove the most manual work, which vendor can ship the broadest set of features. Those questions are legitimate, but they are incomplete. A model can be powerful and still be difficult to govern. A workflow can be efficient and still be impossible to audit. A system can generate useful recommendations and still leave an organization unable to explain how those recommendations shaped a final decision.

This is where the Babel metaphor becomes operational.

A company may introduce AI into recruiting to summarize candidates, prioritize applications, or draft interview notes. Another may use AI in financial services to help analyze creditworthiness, detect risk, or produce customer-facing explanations. A public institution may use AI to support access to benefits or manage citizen requests. In each case, the question is not only whether the system performs well in a benchmark or saves time for the team using it. The deeper question is whether the organization can still explain the relationship between the AI output and the human decision that follows.

When that relationship becomes unclear, the organization has not necessarily suffered a technical failure. The system may be producing answers, routing work, and accelerating decisions exactly as designed. The problem is that understanding has become fragmented across prompts, integrations, logs, documents, vendor systems, and informal human judgment. Different teams may still be speaking the language of efficiency, but they no longer share the same language of responsibility.

That is the modern Babel problem.

The Vatican’s AI statement puts the human person at the center

One of the strongest parts of Magnifica Humanitas is that it refuses to treat artificial intelligence as a purely technical achievement. The encyclical recognizes that AI can be useful, powerful, and beneficial, but it also insists that machine intelligence should not be confused with human judgment.

AI systems process data, identify patterns, generate language, and simulate forms of reasoning. They can produce text that sounds empathetic, legalistic, analytical, or authoritative. But they do not live inside human relationships. They do not carry moral responsibility. They do not understand vulnerability, forgiveness, dignity, or the possibility that a person may change.

That distinction becomes critical when AI is used in decisions that affect people’s lives.

The Vatican statement points directly to sensitive areas such as employment, credit, public services, and reputation. These are not neutral administrative categories. They shape whether a person gets a job, receives financial support, accesses an essential service, or is treated with trust by an institution. When AI enters those workflows, the issue is not simply whether the model can classify, rank, summarize, or recommend. The issue is whether a human institution remains responsible for the judgment being made.

A machine may help analyze information about a person, but it cannot become the moral authority that judges the person. That is the line enterprises need to protect.

This does not mean AI has no place in serious workflows. It means that the more consequential the workflow becomes, the more important it is to preserve human responsibility around it. A person affected by an AI-assisted decision should not be trapped inside an automated process that no one can explain. They should be able to understand why a decision was made, how to challenge it, and which human authority can be held accountable for it.

The EU AI Act turns the same concern into regulation

The Vatican gives this concern moral language, while the EU AI Act gives it a regulatory form.

The European Commission classifies certain AI systems as high-risk when they may affect health, safety, or fundamental rights. That includes AI used in areas such as education, employment, access to essential public and private services, credit, law enforcement, migration, justice, and democratic processes.

The obligations attached to high-risk AI systems are revealing because they do not focus only on model performance. They also address risk management, data quality, logging, documentation, transparency for deployers, human oversight, robustness, cybersecurity, and accuracy. In practice, the regulation is asking whether the system around the model can be governed, not just whether the model can generate an answer.

That is an important lesson for enterprise leaders.

Human oversight cannot be reduced to placing a person somewhere near an automated workflow. A manager who approves an AI-assisted decision without understanding what data was used, what model produced the recommendation, what constraints were applied, or what alternatives were considered is not truly exercising oversight. They are giving institutional legitimacy to a process they may not be able to evaluate.

This is why “human in the loop” is often too weak as a governance concept. The real standard should be whether the human has enough context, authority, and visibility to make a meaningful judgment. If the system hides the relevant information, if the workflow moves too quickly for review, or if the AI output becomes practically impossible to challenge, human oversight becomes ceremonial.

The EU AI Act and the Vatican statement arrive from very different traditions, but they point toward the same enterprise reality: when AI affects people, organizations need more than good intentions. They need systems designed so that responsibility remains visible.

Human judgment needs infrastructure

Many organizations begin responsible AI conversations with principles. They write policies, form committees, define acceptable use guidelines, and tell employees to keep humans involved in important decisions. All of that is necessary, but it is not enough once AI moves from experimentation into daily operations.

Human judgment depends on the technical and organizational environment around it. A person can only remain accountable for an AI-assisted decision if the organization can reconstruct how that decision was produced, including where the AI system operated, what information it was allowed to access, which model or models were involved, how the request moved through the workflow, what output was generated, and what human review took place before the output became action.

Without that surrounding architecture, human oversight becomes fragile. People may be asked to approve decisions that have already been shaped by systems they cannot inspect. Security teams may be expected to manage risk without visibility into data flows. Compliance teams may be asked to defend processes that were never designed to be auditable. Executives may be told that AI is improving productivity while losing the ability to explain how that productivity is being achieved.

This is why the next phase of enterprise AI will not be defined only by better models. It will be defined by governable systems.

At Zylon, we believe organizations should be able to adopt AI without losing control over the infrastructure that shapes their work. That starts with deployment. A private enterprise AI platform, such as Zylon’s platform, gives organizations a way to run AI inside their own environment instead of pushing sensitive workflows into systems they cannot fully control.

That matters because governance is much harder when the organization cannot clearly define where data is processed, how access is managed, which knowledge sources are available, and how AI usage is audited. In regulated industries, those questions are not technical details. They are part of the institution’s duty of care.

The opposite of Babel is not fear

The Vatican’s AI statement does not argue that technology should be rejected. It recognizes that technology can help heal, educate, protect, connect, and expand human possibility. The question is not whether organizations should build with AI. The question is whether they can build in a way that preserves human dignity, responsibility, and shared understanding.

That is why the encyclical’s contrast between Babel and Nehemiah’s rebuilding of Jerusalem is useful. Babel represents a project that grows through uniformity and ambition. Nehemiah represents reconstruction through participation, responsibility, and attention to the human community being rebuilt.

There is a direct enterprise parallel.

Responsible AI adoption should not mean imposing one monolithic system across every team and every workflow. Legal, engineering, finance, customer support, HR, and compliance do not all work with the same data, the same risks, or the same obligations. A low-risk summarization task does not require the same governance model as an AI-assisted workflow that affects employment, credit, legal rights, or access to essential services.

The goal is not to make every team use AI in exactly the same way. The goal is to let teams use AI within boundaries that match the sensitivity of the work.

That requires more than a chatbot interface. It requires model control, data boundaries, permissions, auditability, and deployment options that reflect the organization’s risk profile. With Zylon AI Core, private AI can sit inside the organization’s environment. With the Zylon API Gateway, teams can build AI applications and automations while security and compliance leaders retain controls such as authentication, authorization, model access policies, guardrails, rate limits, and audit logging.

This is what human-centered AI looks like when it becomes operational. It is not a slogan about ethics, and it is not a checkbox added at the end of procurement. It is an architecture that allows people to understand and govern the systems they rely on.

Do not build systems no one can answer for

The public debate around AI often gets pulled toward questions about whether machines can think, reason, or eventually surpass human intelligence. Enterprises have to ask a more immediate question: when AI becomes part of institutional decision-making, can the organization still explain and own the decisions made in its name?

That question becomes unavoidable as AI moves from isolated productivity use cases into workflows that affect employees, customers, citizens, patients, suppliers, and partners. A company that uses AI to draft an email is facing a different level of risk than a company using AI to influence hiring, credit, compliance, or access to services. As the stakes rise, the organization must be able to explain why AI was used, what information shaped the output, how the output was reviewed, and what path exists for correction when the system gets something wrong.

These are not abstract philosophical concerns. They are procurement concerns because companies need to understand what they are buying. They are security concerns because sensitive data may be exposed or misused. They are compliance concerns because regulation increasingly expects traceability and oversight. They are leadership concerns because an organization that cannot explain its AI systems cannot fully govern them.

The Vatican’s AI statement is valuable because it gives leaders a deeper vocabulary for what is at stake. It reminds us that the danger of Babel was not ambition alone, but ambition detached from humility, accountability, and the human relationships that make shared work possible.

Enterprise AI has its own version of that temptation. A single platform promises efficiency across every department. A single model becomes the default voice of the organization. A single automated workflow quietly influences decisions that used to require human deliberation. Over time, the language of productivity can hide a deeper loss of accountability.

That is not responsible adoption. It is Babel with better interfaces.

The better path is not anti-AI. It is AI built for human responsibility: systems that can be inspected, governed, constrained, audited, and challenged when necessary. The organizations that succeed with AI will not simply be the ones that build fastest. They will be the ones that can still understand what they have built, explain who it serves, and remain accountable when it speaks.

Sources

• Pope Leo XIV, Magnifica Humanitas: “On Safeguarding the Human Person in the Time of Artificial Intelligence”

• European Commission: AI Act — Shaping Europe’s Digital Future

Author: Cristina Traba Deza, Product Designer at Zylon
Published: May 2026
Cristina designs secure, on-premise AI platforms for regulated industries, specializing in enterprise AI deployments for financial services, healthcare, and public sector organizations requiring full data control, governance, and compliance.