“Computer says no.” This phrase, voiced with devastating effect by David Walliams in the role of a dead-pan petty bureaucrat in the BBC TV Sitcom, Little Britain (2003), is now mimicked by the algorithm granted authority by the Australian Department of Health, Disability and Ageing to establish who is worthy of receiving taxpayer-funded inhouse aged care. The manifest inadequacies of this machine decision-making has recently been exposed by the ABC’s Four Corners documentary. The algorithm, we learn, does not take account of stories an elderly person might provide about their situation, or notes made about that individual. All that matters are Yes/No assessments for a predetermined set of questions. The ghost of Robodebt remains.
These changes in healthcare administration are another manifestation of a profound shift taking place provoked by the dramatic growth of Artificial Intelligence—namely the increased tendency for people to trust a machine’s authority and capacity to respond correctly to the questions we put to it. A Google enquiry provides an AI answer to any question in an authoritative tone, much like a governmental edict, as if it were the omniscient voice of God.
As a historian of medieval Christianity, interested in the sacred stories and songlines of many traditions, including those from pre-conquest Australia, I observe parallels in the way sacred stories about healing in the face of tragedy and injustice evolve when appropriated to serve institutional, and sometimes personal, ends. Such distortion is a form of cultural theft and can be seen in the way AI has enabled a proliferation of scammers who use AI to assert authority by reproducing fake images of trustworthy people, whether in politics, business or medicine.
AI & sacred texts
Equally disturbing is the phenomenon of global AI data storage companies buying up books, ripping pages from their spines, and scanning their contents, so as to own texts as data, maximised for the financial profit of a few TechBros. Legions of lawyers are employed to resist the efforts of authors, musicians and visual artists to maintain traditional copyright agreements. What then is to stop generative AI swallowing up sacred texts and re-composing a new narrative?
I was recently shown an AI-generated sacred poem Lady Poverty—a central figure in the Franciscan tradition. The poem is set as a dialogue in rhyming couplets between St Francis (1181/2-1226) and St Clare (1194-1253) and crudely amalgamates some wonderful narratives about the friendship between these two remarkable individuals and the personification of Lady Poverty. The poem also draws on the Sacrum Commercium (Sacred Exchange), a fascinating anonymous narrative about St Francis’s relationship with Lady Poverty. This important text critiques the institutional Church as failing to embrace poverty. The AI-generated poem parrots authentic Franciscan phrases with banal rhymes (“Lady Poverty, gentle bride, / Keep me close to Jesus’ side”), misrepresenting core Franciscan values.
AI is of course not the first “author” to give new meanings to sacred texts or values. The so-called Protevangelium of James (Gospel of St James) was compiled in the second half of the second century from material in all four Gospels. It became a key text in the worship of Mary and is the origin of the idea of “perpetual virginity”. The Protevangelium introduced Mary’s parents, Anne and Joachim, and described their unexpected conception of Mary. It also focussed on Mary giving birth to Jesus while miraculously preserving her perpetual virginity. By paraphrasing existing scriptures, it became a hugely influential sacred story that elevated cultic virginity above the state of marriage. In the late fourth century, after St Jerome (c. 342-347) accepted this teaching, many people assumed it to be God’s voice alongside the other Gospels. Protevangelium’s authority endured until the Reformation, when marriage became sanctioned as an ideal. This elevation of marriage failed to recognise that some women rejected marriage because of the physical danger of childbirth.
AI, parrots & data sovereignty
The history of Protevangelium tells us that the “recreation” of sacred texts can change the shape of society and faith in fundamental ways. There is no reason to think that AI cannot do the same. In 2021 Emily Bender et al. formulated the brilliant image of a stochastic (i.e. random) parrot to show how AI mimics existing phrases to create seemingly authoritative answers.
While modern manipulations inevitably lack the literary sensitivity of Greek or Latin retellings of existing narratives, the process of recombining existing phrases to create “new” texts is not so unusual. Readers need to approach these texts with a critical eye lest they be misled and simply accept AI answers at face value as “authoritative”, as they are retellings that often serve a particular, but hidden, agenda.
The question of how to deal with Indigenous sacred songs and stories, and the desire to protect Indigenous sacred texts from extractive scholarship has led many repositories to apply concepts of Indigenous Data Sovereignty and introduce mediated access requirements, set out to ensure that communities maintain access to records of Indigenous teaching, and authority over their use. There is currently active debate on restrictions placed on broader scholarly access to these records and concern, in some cases, we are yet to get this balance right. AI’s appropriation of human writing brings complex notions of ownership and sovereignty over texts and images into sharp relief within a history of the use and misuse of sacred or treasured knowledge.
As any Latin student knows, the word “data” means “things given”. Religious texts are themselves “data”, requiring a personal response in their receipt. When a health bureaucrat examines the data collected from an elderly person, there should be an expectation of reciprocity. Positivist assumptions about the objectivity of “evidence-based claims” fail to recognise that all our understandings are based on stories “given” to another person. AI is simply a tool; it is not an authority or a receiver capable of “reciprocal reciprocity”. As Pope Leo XIV argues in his encyclical Magnifica humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence, it is up to us to ensure we are the ones who ask critical questions, and not blithely accept the answers devised by an algorithmic system that likes to think of itself as omnipotent and enriches only a few. Just as St Francis challenged the image of the Church by promoting Lady Poverty rather than cultic virginity, so we must challenge those claiming to speak with the voice of God by finding songs of wisdom in sacred stories rather than blindly repeated scripts of hate and blame.
