Magisterium Ai

Matthew Harvey Sanders, founder of Magisterium AI, tells AdVaticanum that Pope Leo XIV’s Magnifica Humanitas has set a new benchmark for how Catholics should approach artificial intelligence, warning against treating AI as if it possesses conscience or moral interiority The founder of the Catholic artificial intelligence platform Magisterium AI has said that Pope Leo XIV’s first encyclical on artificial intelligence has set a new benchmark for how Catholic institutions should approach emerging technologies, while warning against treating AI systems as if they possess conscience or moral interiority. Matthew Harvey Sanders, founder and chief executive of the Canadian technology company Longbeard, spoke exclusively to AdVaticanum following the publication of Magnifica Humanitas , the Pope’s inaugural encyclical, which was formally presented at the Vatican this week on AI. Matthew Sanders described the atmosphere surrounding the launch as unlike anything he had experienced in the technology sector, saying the event brought together Vatican officials, clergy, engineers, investors and researchers in a setting marked by “a particular solemnity”. “I have attended a great many events in the technology sector,” he said. “The hall where Leo XIV presented Magnifica Humanitas did not feel like any of them.” He said Pope Leo approached the subject of artificial intelligence without either fear or fashionable enthusiasm, adding that the Pontiff appeared entirely comfortable discussing the implications of rapidly developing technologies. “He spoke without the defensive caution you often find from institutional leaders engaging with technical topics, and without the performative enthusiasm of someone trying to signal relevance,” Sanders said. “He was simply present to the conversation.” Sanders argued that the most significant aspect of the papal intervention was not a condemnation of artificial intelligence itself, but what he described as an “invitation” by the Pope to technology developers and researchers to engage seriously with questions concerning the human person. “What he extended that day was an invitation to every laboratory and every developer in the field: not a verdict on the technology, but an ongoing engagement about what it means for the human person,” he said. “That invitation, in my view, is the most consequential thing that could have come out of the day.” Sanders also disclosed that he spent time during the Vatican gathering speaking with two senior researchers from the American artificial intelligence company Anthropic, including Chris Olah, whose work has focused on mechanistic interpretability, and Amanda Askell, who leads research into the behavioural character of the Claude AI model. “Chris Olah has spent years on mechanistic interpretability: the painstaking effort to reverse-engineer a trained neural network and understand, at a granular level, what is actually occurring when the model processes language,” Sanders said. Referring to Askell, he added: “Amanda Askell leads the work on Claude’s character; she has conducted more careful, sustained inquiry into how a large language model behaves under pressure, across context and at the edges of its training than arguably anyone working today.” Sanders said that while both conversations were “genuinely interesting”, he nevertheless believed the Pope was correct to reject attempts to describe artificial intelligence systems as possessing conscience or moral interiority. “There is an openness in parts of the research community to describing current models as possessing something resembling conscience or moral interiority,” he said. “The encyclical addresses this head-on, and I think the Pope is right to resist it. “Attributing moral subjecthood to a statistical system is a category error with consequences that go well beyond the lab.” Sanders said the presence in the same room of leading AI researchers alongside the Bishop of Rome reflected the scale of the questions now confronting both the Church and the technology industry. “The fact that the conversation in that room included both of them and the Bishop of Rome struck me as exactly the kind of encounter this moment requires,” he said. According to Sanders, Pope Leo’s principal concern is not simply what legal or regulatory restraints should govern artificial intelligence, but what understanding of the human person should precede and shape those rules. He connected the encyclical’s arguments to themes he explored earlier this year in an essay entitled “The Church as the Ark for a Post-Work World”, in which he argued that the greatest disruption caused by artificial intelligence would not ultimately be economic but existential. “My argument there was that the coming crisis of AI is not fundamentally economic: it is existential,” Sanders said. Referring to the Austrian psychiatrist Viktor Frankl, he added: “Viktor Frankl described what happens when a civilisation’s anchor for human identity is removed: an ‘existential vacuum’, a suffocating meaninglessness that no material provision can address.” “Silicon Valley’s answer to the disruption of labour is what I called the ‘hollow utopia’: income to fund the body, and infinite digital distraction to occupy the mind,” Sanders said. He contrasted this with the vision presented in Magnifica Humanitas , arguing that the Pope had rejected the idea that efficiency or economic productivity alone can define human flourishing. “When he cites John Paul II in §129, asking whether AI makes human life on earth ‘more human’ in every aspect of that life and more worthy of man, he is insisting the criterion is not comfort or efficiency,” Sanders said. He also pointed to another section of the encyclical which states that “no computational system, however sophisticated, can create a heart that gives itself, or a conscience that discerns good from evil”. “The GDP era told us human value was about output,” Sanders said. “The encyclical says it never was.” Speaking about the implications of Magnifica Humanitas for Magisterium AI, Sanders said he wanted “to be careful not to appropriate it for our own promotional purposes”, but argued that the encyclical nevertheless addressed “something much larger than any single company”. He highlighted the line: “A more moral AI is not enough if that morality is determined by a few”, adding: “That sentence captures the structural reason Magisterium AI was built the way it was built.” Sanders insisted that “a general-purpose AI platform cannot be made Catholic simply by pointing its outputs at Catholic content”, because “the moral architecture of a system, who controls its training data, what its reward functions are optimised for, what the company that built it needs commercially, shapes every response it generates”. He added that Magisterium AI was built around what the company calls the “off-ramp”, explaining: “Catholic AI should answer a question and then send the person back into prayer, real relationships and the sacramental life of the Church. It should be designed to be finished with.” He also cited another line from the encyclical stating that “technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate and use it”. Referring to the encyclical’s treatment of subsidiarity, he warned that many Church institutions presently rely on technologies controlled by corporations whose priorities may change without notice. “They are using tools built and governed by a handful of companies, running on infrastructure they neither own nor fully understand,” he said. Sanders recalled advice once given to him by the former Archbishop of Toronto, Cardinal Thomas Collins. “My former boss, Cardinal Thomas Collins, used to say: ‘If you know where you’re going, you’ll be more likely to get there.’” Near the conclusion, Sanders said: “The encyclical has sharpened the question of where Catholic institutions should be going with this technology. It has not done the building for us. “The work of constructing digital infrastructure that actually embodies these principles is still largely ahead of us. But the standard has been set.”