In a striking display of convergence between ancient faith and cutting-edge technology, Pope Leo XIV recently unveiled his first encyclical on artificial intelligence, Magnifica Humanitas. The Vatican event featured a notable guest: Christopher Olah, cofounder of Anthropic, an AI safety company. This unprecedented alliance signals a powerful new chapter in the global conversation surrounding artificial intelligence, bridging the moral authority of the Catholic Church with the innovative spirit of Silicon Valley.
To truly grasp the significance of this partnership, we must look back to Anthropic’s origins. The company emerged in 2021 from a pivotal moment, when a group of OpenAI researchers, including siblings Dario and Daniela Amodei, departed to form a rival laboratory. Their motivation was clear: they believed that AI models were rapidly becoming too potent to be developed solely under the pressures of competition and speed.
The Genesis of an Ethical AI Movement
Since its inception, Anthropic has meticulously built its public identity around the crucial concept of AI safety. Their mission extends beyond merely creating powerful models; they strive to develop systems that are both controllable and inherently guided by ethical principles. This foundational philosophy gave rise to their innovative approach known as Constitutional AI.
Constitutional AI involves training systems not through endless manual corrections of undesirable outputs, but by imbuing them with a foundational “constitution” of principles and rules. This method aims to align AI behavior with human values from the ground up, moving beyond reactive fixes to proactive ethical design. It’s a deliberate effort to ensure AI systems understand and adhere to a moral framework.
The Vatican’s Proactive Stance on AI Ethics
Christopher Olah’s presence at the Vatican was no mere coincidence or last-minute symbolic gesture. It represented the culmination of a deliberate, long-term strategy by the Holy See to evolve from a distant moral observer of technology into a direct and influential interlocutor within the AI industry. The Vatican recognized the profound implications of this rapidly advancing field.
A significant initial stride in this direction was the Rome Call for AI Ethics in 2020, an initiative spearheaded by the Pontifical Academy for Life alongside tech giants like Microsoft and IBM. This landmark collaboration aimed to establish universal ethical principles for AI development, advocating for essential values such as transparency, inclusion, and accountability. These principles formed a crucial groundwork for future engagements.
While the Vatican initially focused on bioethics, the landscape quickly transformed with the ascent of models like ChatGPT, the intensifying global tech race, and the escalating power of Big Tech. These developments convinced the Holy See that the discussion had transcended mere technical ethics, now touching upon the fundamental future of humanity itself. In this evolving context, Anthropic, with its dedicated focus on AI safety, naturally emerged as a pivotal partner.
Bridging Faith and Technology: Shared Concerns
The Vatican has closely monitored a specific facet of the AI debate: the critical issue of AI model alignment. This is precisely where Christopher Olah’s expertise becomes indispensable. Unlike the more public-facing Amodei siblings, Olah embodies the deeply theoretical and philosophical dimension of AI research, making him a leading voice in the field.
Olah is globally renowned for his work on model interpretability, striving to demystify the inner workings of increasingly complex neural networks. On his personal website, he articulates his goal as “transforming neural networks into algorithms understandable to human beings.” This objective perfectly resonates with the core concerns of Pope Leo XIV’s encyclical, which deeply reflects on the dangers of creating technologies too powerful to be fully comprehended, controlled, or governed by humanity.
Reports suggest that the interactions between the Vatican and Anthropic intensified during various global summits on AI safety. The Holy See recognized Anthropic as a company willing to openly acknowledge that the profound challenges posed by artificial intelligence cannot be resolved by the technology industry in isolation. This shared understanding forms the bedrock of their collaboration.
Both the encyclical and its presentation emphasized a crucial point: technology is never truly neutral, and algorithms inherently embody a particular worldview. Anthropic’s Constitutional AI project directly addresses this by explicitly integrating human values, rules, and ethical principles into the operational behavior of its AI models. This proactive approach seeks to ensure AI development serves humanity’s best interests.
Ultimately, the connection between the Vatican and Anthropic is rooted in a shared apprehension: the fear that increasingly powerful AI systems could be shaped solely by the economic, geopolitical, and competitive pressures dominating the global AI race. This common concern about unintended consequences unites their disparate institutions. For Anthropic, this relationship also carries immense reputational value, positioning it as a leader in ethical AI at a time when technology’s societal impact is under intense scrutiny across labor, national security, and surveillance domains.
“Magnifica Humanitas”: A Call for Human-Centered AI
Anthropic’s chatbot, Claude, is a testament to this philosophy, built explicitly around the concept of trust with users. Its responses are guided by an ethical constitution, and the company’s public discourse consistently revolves around responsibility and safety. This strategic emphasis ensures that ethics is not just an afterthought but an intrinsic part of the product’s symbolic infrastructure and brand identity.
The very title of Pope Leo XIV’s encyclical, Magnifica Humanitas, encapsulates a profound tension: humanity is described as magnificent, yet simultaneously capable of creating new forms of dehumanization. From this perspective, artificial intelligence is not intrinsically evil; rather, it functions as a powerful mirror reflecting the values and intentions of its creators. The technology itself is a tool, its impact determined by human choices.
The Vatican delivers a stern warning against the emergence of a new “digital Babylon,” a society where people, relationships, and even truth itself are reduced to mere data, performance metrics, and efficiency targets. This prophetic vision cautions against a future where human dignity is sacrificed at the altar of technological advancement. The encyclical explicitly critiques the growing concentration of technological power in the hands of a limited number of transnational private entities, echoing contemporary debates on AI governance.
Key questions arise: Who truly controls these advanced AI models? Who dictates the criteria by which they are trained and developed? Who will ultimately own the foundational infrastructure of our technological future? During the encyclical’s presentation, Christopher Olah made a remarkably candid admission for a tech executive, acknowledging that even ethically minded companies remain deeply influenced by economic and competitive pressures that can conflict with “doing the right thing.”
Olah’s statement publicly reinforced the Vatican’s core message: the complex challenges of artificial intelligence transcend self-regulation by technology companies alone. Comparisons to the atomic bomb are frequent in AI discussions, but a critical distinction exists: nuclear technology was predominantly state-controlled, whereas AI development largely resides within private corporations. This pivotal point is central to the encyclical, highlighting that today, “technological power takes on a new face, one that is predominantly private.”
This reality underpins a shared, profound fear, articulated in different languages by the Vatican and by AI safety proponents: that increasingly powerful systems could be guided by distorted human incentives. The “Hiroshima of the 21st century,” as cautioned by the encyclical, may not be a single, catastrophic event, but rather a gradual process of social automation. In this scenario, humans slowly delegate their critical thinking, decision-making, information gathering, and relational capacities to machines. The risk, therefore, is that our Magnifica Humanitas could subtly, but irrevocably, transform into something truly terribilis.
Source: Wired – AI
