CATHOLIC Ai

To ensure a meaningful public debate on the capabilities, risks and regulation of AI, we should strive to use more nuanced terminology.

This applies in particular to journalists, who shape public perception and establish new terminology.

Magnifica Humanitas,

the encyclical that Pope Leo XIV released last May and that represents the most articulate and authoritative teaching on the topic.

Antiqua et Nova,

a note published by the Dicastery for the Doctrine of Faith and the Dicastery for Culture and Education in January 2025 on the relationship between artificial intelligence and human intelligence.

Quo Vadis Humanitas?,

a document of the International Theological Commission published in February 2026 that contains an extensive reflection on what theological anthropology has to say to the challenges of our technological age.

Message for the 60th World Day of Social Communications,

a succinct but comprehensive address in which Pope Leo previewed many of the encyclical’s contents.

60 years ago it ment electronic hardware, 50 years ago it was hardware and software, today it includes all this and anything that has a “screen”

This is a very important question. Talk of “artificial intelligence” is as ubiquitous in today’s discourse

AI “does not refer to a single technology—it is an umbrella term” that is not new. It can name an entire field of knowledge and computer science research that stretches back to the 1950s devoted to creating machines that would exhibit abilities that we typically associate with human intelligence.

In this broader sense, artificial intelligence encompasses all sorts of software and applications that predate the more recent advent of ChatGPT and its competitors and that are fully integrated and necessary to our contemporary technological society.

Artificial intelligence can also be a way to refer to machine learning, that is, the creation of models that use algorithms “to make predictions or decisions based on data.”

Even in this case, to use artificial intelligence to refer to machine learning is a way to refer to all sorts of models, softwaree, and applications, some of them quite dated.

Recent advances in computational power and breakthroughs in algorithmic techniques have made machine learning more advanced and powerful than ever before, giving rise to what experts in the field call “deep learning,” a type of machine learning that allows models to identify complex patterns and relationships in large amounts of data without constant and direct human intervention.

It is from this kind of machine learning that generative AI comes. IBM defines generative AI’s distinctive mark as the ability to create “complex original content such as long-form text, high-quality images, realistic video or audio and more in response to a user’s prompt or request.” It does so by drawing on extensive amounts of data that constitute its training and producing original content that draws on the patterns and relationships the machine detected in the data used in its training. 

Steinbrück describes large language models (LLMs) as a form of generative AI that have learned “how human language works by analysing vast amounts of text” and can thus produce highly sophisticated outputs based on them via a process called tokenization. These LLMs are what power the chatbots we have come to know (ChatGPT, Claude, Gemini, and the like), and they are what most people refer to when they think of artificial intelligence.

LLMs and generative AI are certainly what the Church’s documents listed above have in mind when wrestling with the opportunities and challenges posed by artificial intelligence. Yet, the technology has not stopped there. What Ethan Mollick helpfully describes as “conversation-first” AIs that depend on users constantly interacting with them by asking questions have now been surpassed by so-called “AI agents,” goal-directed AI systems that, as IBM explains, “can perform tasks and accomplish goals on behalf of a user or another system without human intervention, by designing [their] own workflow and using available tools (other applications or

services).” Steinbrück explains that AI agents are “equipped with a kind of ‘executive’ capability, meaning they can not only generate text but also access operating system functions or control entire user accounts.” The user gives them goals, rather than step by step instructions, and the AI agents use tools, memory, and external systems to respond to the user’s query. This is the new AI frontier, one that is so recent that, despite being published just three months ago, Magnifica Humanitas does not address directly, although it offers principles for discernment and action that can be easily extended to this latest form of AI technology. The complexity and constant evolution of the multiple technologies that fit under the umbrella term artificial intelligence is one of the reasons why the Catholic Church calls for principled and careful discernment and does not simply arrive at definitive conclusions in favor or against AI - the notable exception being its non-negotiable opposition against fully autonomous weapons systems (Magnifica Humanitas, no. 200).