The first Catholic Universities Art Triennial seeks to intervene in a question that already overflows laboratories and technology companies: what is worth preserving when a machine can produce images, texts, and sounds difficult to distinguish from those created by a person.
The project is called ARTCUT. It is promoted by the Holy See's Dicastery for Culture and Education, with Nuno Crespo as chief curator. Its first edition, Exercises in Empathy, takes place between August and December 2026 at Catholic universities in Rio de Janeiro, Santiago, Bogotá, Boston, Milan, Lisbon, and Luanda. The programme will culminate in Venice. The list brings together artists, curators, researchers, and institutions from very different traditions; it is not an exhibition conceived around confessional art.
The presentation document links the triennial to the debate on artificial intelligence. It cites the formulation of Leo XIV according to which these systems can simulate empathy and understanding, but do not understand what they produce because they lack the affective, relational, and spiritual perspective from which human beings grow in wisdom. ARTCUT proposes to delve into that debate and "safeguard what is fundamentally human".
The question deserves attention. The Church's response, however, gathers distinct arguments: a theological anthropology, an ethics of responsibility, and a definition of art. The first two can be sustained without the third becoming inevitable.
The Church's position is broader than a defence of authorship
The note Antiqua et nova, published on 28 January 2025 by the Dicastery for the Doctrine of the Faith and the Dicastery for Culture and Education, is not limited to art. It distinguishes human intelligence from the functional capacity of AI systems. The former belongs to an embodied, historical, and relational person; the latter process data and perform tasks within a logical-mathematical framework. The document acknowledges that AI can be trained on products of human creativity and generate artefacts difficult to distinguish from a human composition. The similarity of the result does not, in its approach, eliminate the difference between a person and a tool.
That distinction has solid ethical consequences. Responsibility cannot be delegated to a model. The one who designs, contracts, deploys, selects, and uses the system is accountable. The note also points to the concentration of power in a few companies, the extraction of value, the potential degradation of work, automated surveillance, and the reduction of relationships to simulated exchanges. None of that depends on deciding whether an algorithmic output is art.
The document takes an additional step when it attributes to human intelligence an openness to truth, goodness, and beauty that the machine does not possess. In the address of 2 October 2026 to artists and curators, Leo XIV converts this anthropology into an artistic boundary: he asserts that there is an ontological difference, prior to the aesthetic difference, between human art and that which a machine generates through statistical calculation on millions of others' images.
The statement gathers three problems that deserve to be separated:
- The provenance and licensing of training data.
- Human responsibility for the use of the models.
- The artistic status of a configuration produced with them.
The first two are legal, economic, and ethical. The third is ontological. From the fact that a machine is not a person, does not feel, and does not respond morally, it does not follow that no configuration in which it intervenes can produce an artistic event.
Empathy, understanding, and the work
Exercises in Empathy adopts a more fertile title than the opposition between human production and automatic production. Empathy is not a substance lodged within the author. It is a relationship, a practice, and, at times, an effect. It can exist in the production process, in the interaction between participants, in the manner of exhibition, in the visitor's attention, or in the consequences of the work. It can also be absent in a work made entirely by a person.
An author may feel intensely and produce an inert configuration. Another may work with rules, chance, delegation, or machines and open a relationship that no one had foreseen. Private emotion also does not guarantee that a work cares for those who receive it. The history of art contains objects of enormous aesthetic efficacy conceived to humiliate, exclude, discipline, or glorify violence.
AI does not experience empathy. A calculated response accompanies no one and contracts no obligations. Even so, a work does not have to possess the faculties of a person. Neither does a carved stone, a musical score, a film, or an instruction executed by another person. The artistic question begins when a configuration enters a field of relationships and modifies its possibilities.
This does not equate a model with an artist. It shifts the question. Instead of deciding in advance what can be art according to the nature of who or what intervenes in the production, it examines what the work reorganises, under what conditions, and with what human responsibilities.
My research on Structural Surplus formulates that criterion as structural surplus: a relational difference that effectively and non-trivially persistently reorganises the field of possible configurations of a system, without being absorbed by its primary function. Consciousness is the dominant historical form of artistic production, but it does not function as a necessary condition for every artistic event.
Art as a search engine
Leo XIV's discourse contains an image more productive than its ontological frontier. A search engine indexes what exists and tends to return what we already expect. Art, he says, is another type of "search engine": it does not proceed through indexing, but through interrogation. It resists flattening, slows down the gaze, and forces one to pause before that which does not allow itself to be consumed.
That criterion does not ask first about the producer's biography. It asks what operation the work performs. The relevant opposition appears between confirmation and interrogation, between a difference that the system absorbs and one that alters its possibilities.
Most AI-generated images do not pass that test. They optimise recognisable codes, mix available repertoires, and return a variation compatible with expectations. Their statistical novelty can coexist with profound cultural conformity. Something similar happens with an enormous amount of human production. A person's signature does not turn every result into an interrogation.
The issue with generative models extends beyond their lack of interiority. Their industrial architecture favours repetition: datasets converted into regularities, interfaces oriented towards satisfactory responses, opaque safety filters, reliance on providers, and recommendation systems that reward the immediately recognisable. When the model disappears addresses another consequence of this infrastructure: a work may retain its code while losing that which responded behind an API.
The defence of the human gains precision when directed at these material conditions. It is necessary to protect attribution, consent, remuneration, the capacity to challenge, cultural diversity, and control over tools. 'Made by a person' proves an overly impoverished criterion if that person works within a system that denies them time, rights, and agency.
What a triennial on empathy should examine
A university triennial can avoid two frequent reductions. The first consists of using works to illustrate a pre-determined doctrinal conclusion. The second turns artificial intelligence into a technological spectacle while leaving its conditions of production intact.
Exercises in Empathy would be more incisive if it treated empathy as a verifiable practice. Some possible questions:
- Who collected and labelled the materials used by a work.
- Which individuals or communities are represented and whether they were able to decide how.
- What labour relations sustain the production and installation.
- What data is sent to external providers.
- Which licences permit the reuse of images, voices, or texts.
- Who preserves the work when the model or the API disappears.
- How authorship, recognition, and remuneration are distributed.
- What changes in the visitor beyond identifying a compassionate intention.
Documentation of the process does not, by itself, resolve the status of a work, but it allows for a discussion of the responsibilities that accompany it. The guide to documenting the creative process explains how to preserve sketches, versions, metadata, decisions, and proofs of authorship. In a practice involving AI, one would have to add models, versions, parameters, instructions, licences, providers, and selection criteria.
The triennial also possesses a rare advantage: seven university venues in distinct cultural contexts. Empathy could be examined by comparing how a work changes as it circulates between Rio de Janeiro, Santiago, Bogotá, Boston, Milan, Lisbon, and Luanda. If the same proposal produces different relations, the event was not enclosed within the author's consciousness or the digital archive. It emerged from the encounter.
Safeguarding capacities, not a closed essence
The expression 'safeguarding the human' can protect a dignity that does not depend on performance. Antiqua et nova rightly insists that the value of a person does not derive from their cognitive, productive, or technological capabilities. This idea prevents measuring human beings against a machine and considering those who perform less as expendable.
Applied to art, the same formula runs another risk: converting 'the human' into a stable property that institutions administer and certify. History shows that the boundaries of the fully human have excluded bodies, peoples, trades, and forms of knowledge. A defence of humanity needs to acknowledge this history and not use ontology as a cultural customs house.
Safeguarding the human could mean something else: protecting the conditions to question, dissent, care, remember, assume responsibility, and transform the relations we inhabit. These are decisive human capabilities. They do not require asserting that every artistic event begins and ends in an individual interiority.
The triennial is correct to identify homogenisation as a threat. It is also correct to present empathy as an exercise. An exercise admits learning, friction, and failure; it does not designate an essence possessed beforehand.
The Church asks what must be preserved when a machine imitates human results. My provisional answer is less reassuring: human responsibility must be preserved without using it to foreclose art.
On the open conversation
If you work in aesthetics, theology, contemporary art, or generative systems and can provide documentation on ARTCUT or discuss this framework, you may write to me. Reasoned disagreement is part of the research.
The author’s relationship to the cases
In Art as Structural Surplus, I defend a thesis contrary to the ontological boundary proposed in the speech by Leo XIV: human consciousness is not a necessary condition for the artistic event. I use artificial intelligence tools in my work and have published 'Having good taste is not creating', where I distinguish between selection, combination, and the reorganisation of the field of possibilities. I maintain no economic or professional relationship with the Holy See, the Dicastery for Culture and Education, or the universities participating in ARTCUT.