Why the Vatican Invited Anthropic to the Pope’s AI Encyclical Presentation

为什么梵蒂冈邀请 Anthropic 参加教皇的人工智能通谕发布会

When Pope Leo XIV presented his first encyclical on artificial intelligence at the Vatican on Monday, he invited Christopher Olah, cofounder of Anthropic, to speak. The move signaled an unprecedented alliance between the Catholic church and Silicon Valley. But to understand how this partnership came about, we need to go back to Anthropic’s founding. 周一，当教皇利奥十四世在梵蒂冈发布其首份关于人工智能的通谕时，他邀请了 Anthropic 的联合创始人克里斯托弗·奥拉（Christopher Olah）发表讲话。此举标志着天主教会与硅谷之间达成了一种前所未有的联盟。但要了解这种伙伴关系是如何形成的，我们需要回顾 Anthropic 的创立过程。

Why Anthropic?

为什么是 Anthropic？

Anthropic launched in 2021 after a group of OpenAI researchers, including Dario and Daniela Amodei, left to form a rival lab. They did so with a clear conviction: Artificial intelligence models were becoming too powerful to be developed exclusively according to the logic of competition and speed. Anthropic 于 2021 年成立，当时包括达里奥·阿莫代（Dario Amodei）和丹妮拉·阿莫代（Daniela Amodei）在内的一群 OpenAI 研究人员离开公司，组建了一家竞争对手实验室。他们怀着一个明确的信念：人工智能模型正变得过于强大，不能仅仅按照竞争和速度的逻辑来开发。

Since then, Anthropic has built its public image around the concept of AI safety. The company aims to build not just powerful models, but ones that are controllable and guided by ethical principles. This is where the concept of Constitutional AI comes from: the idea of training systems using a kind of constitution composed of principles and rules, instead of just manually correcting the most risky and dangerous responses. 自那时起，Anthropic 便围绕“人工智能安全”这一概念打造其公众形象。该公司的目标不仅是构建强大的模型，还要构建可控且受伦理原则指导的模型。这就是“宪法人工智能”（Constitutional AI）概念的由来：其核心思想是利用一套由原则和规则组成的“宪法”来训练系统，而不是仅仅手动纠正那些风险最高、最危险的回答。

Pope Leo XIV attends the presentation of his first Encyclical Letter, Magnifica Humanitas, focused on the rise of artificial intelligence, in the Vatican on May 25, 2026. Photograph: Alberto Pizzoli/Getty Images 2026 年 5 月 25 日，教皇利奥十四世在梵蒂冈出席其首份关注人工智能崛起的人工智能通谕《Magnifica Humanitas》（宏伟的人性）的发布会。摄影：Alberto Pizzoli/Getty Images

How the Convergence With the Vatican Began

与梵蒂冈的交集是如何开始的

Olah’s presence at the Vatican was obviously not accidental, nor the result of a last-minute symbolic gesture. It was the outcome of a deliberate, long-term effort in which the Vatican has progressively sought to transform itself from a moral observer of technology into a direct interlocutor with the AI industry. 奥拉出现在梵蒂冈显然并非偶然，也不是临时起意的象征性举动。这是梵蒂冈长期深思熟虑的结果，梵蒂冈一直在寻求从技术的道德观察者转变为与人工智能行业直接对话的参与者。

The first major step came in 2020 with the Rome Call for AI Ethics, an initiative promoted by the Pontifical Academy for Life together with Microsoft, IBM, and other international organizations. The goal was to establish a shared foundation of ethical principles for the development of AI, including transparency, inclusion, and accountability. 第一个重大步骤是 2020 年的《罗马人工智能伦理呼吁》（Rome Call for AI Ethics），这是一项由宗座生命科学院与微软、IBM 及其他国际组织共同推动的倡议。其目标是为人工智能的发展建立一个共同的伦理原则基础，包括透明度、包容性和问责制。

At the time, the Vatican appeared to be operating primarily in the realm of bioethics and moral questions. In the years that followed, however, the context changed dramatically. The rise of ChatGPT, the struggle for technological leadership between the United States and China, and the growing power of Big Tech gradually convinced the Holy See that the issue was no longer just about tech ethics, but about the very future of humanity. 当时，梵蒂冈似乎主要在生物伦理和道德问题领域开展工作。然而，在随后的几年里，背景发生了巨大变化。ChatGPT 的崛起、美中之间的技术领导权之争，以及大型科技公司日益增长的影响力，逐渐让圣座意识到，这个问题不再仅仅关乎技术伦理，而是关乎人类的未来。

In this sense, Anthropic has come to be seen by the Vatican as a particularly important interlocutor. Unlike other Silicon Valley companies that have built their reputations primarily around innovation and growth, Anthropic has made AI safety a core part of its identity. 从这个意义上说，梵蒂冈将 Anthropic 视为一个特别重要的对话者。与其他主要围绕创新和增长建立声誉的硅谷公司不同，Anthropic 将人工智能安全作为其身份的核心部分。

In recent years, the Vatican has followed one specific strand of the technology debate with particular attention: the alignment of AI models. 近年来，梵蒂冈一直特别关注技术辩论中的一个特定方向：人工智能模型的对齐（Alignment）。

Olah’s Role

奥拉的角色

This is where Christopher Olah comes in. Unlike the Amodei siblings, who are more exposed to the media, Olah represents the more theoretical and almost philosophical side of AI research. He is one of the world’s best-known researchers on the topic of model interpretability, or the effort to understand what really happens inside increasingly complex neural networks. 这就是克里斯托弗·奥拉发挥作用的地方。与更多出现在媒体视野中的阿莫代兄妹不同，奥拉代表了人工智能研究中更具理论性、甚至近乎哲学的一面。他是全球最知名的模型可解释性研究人员之一，致力于理解日益复杂的神经网络内部究竟发生了什么。

Christopher Olah Photograph: Alessia Giuliani/Getty Images 克里斯托弗·奥拉。摄影：Alessia Giuliani/Getty Images

On his personal website, Christopher Olah describes himself as someone trying to “transform neural networks into algorithms understandable to human beings.” And it is difficult to imagine a figure more aligned with the core of Pope Leo XIV’s encyclical: a reflection centered on the risk of building technologies that become too powerful to be understood, controlled, or governed. 在他的个人网站上，克里斯托弗·奥拉将自己描述为试图“将神经网络转化为人类可理解的算法”的人。很难想象还有谁比他更契合教皇利奥十四世通谕的核心：即对构建过于强大以至于无法被理解、控制或治理的技术所带来的风险进行反思。

According to various journalistic sources, the contacts between circles close to the Holy See and Anthropic may have intensified right during the global summits on AI safety. The Vatican saw in Anthropic a company at least willing to publicly acknowledge that the problem of artificial intelligence cannot be solved by the technology industry alone. 据多家新闻来源称，圣座周边圈子与 Anthropic 之间的联系可能正是在全球人工智能安全峰会期间加强的。梵蒂冈在 Anthropic 身上看到了一家至少愿意公开承认“人工智能问题不能仅靠科技行业自身解决”的公司。

This is a point emphasized both in the text of the encyclical and during its presentation. The document repeatedly insists on the idea that technology is not neutral, and that algorithms inevitably embody a particular worldview. Through its Constitutional AI project, Anthropic is attempting to do precisely this: explicitly introducing values, rules, and principles into the behavior of AI models. 这一点在通谕文本及其发布会上都得到了强调。该文件反复坚持认为技术并非中立，算法不可避免地体现了特定的世界观。通过其“宪法人工智能”项目，Anthropic 正试图做到这一点：将价值观、规则和原则明确地引入人工智能模型的行为中。

In essence, the connection between the Vatican and Anthropic lies in a shared fear: that increasingly powerful systems could end up shaped solely by the economic, geopolitical, and competitive incentives driving the global AI race. 本质上，梵蒂冈与 Anthropic 之间的联系在于一种共同的恐惧：日益强大的系统最终可能完全被驱动全球人工智能竞赛的经济、地缘政治和竞争激励所塑造。

Reputation as Product

声誉即产品

The story also has an industrial dimension, not just a spiritual one: For Anthropic, its relationship with the Vatican inevitably carries enormous reputational value. At a moment when artificial intelligence is becoming increasingly central to debates about labor, national security, surveillance, and military power, the image of an “ethical AI company” is a significant advantage. 这个故事不仅具有精神层面，还具有产业层面：对于 Anthropic 而言，它与梵蒂冈的关系不可避免地带来了巨大的声誉价值。在人工智能日益成为劳工、国家安全、监控和军事力量辩论核心的时刻，“伦理人工智能公司”的形象是一个巨大的优势。

Claude, Anthropic’s chatbot, was built precisely around the idea of trust with users. The model “responds” to an ethical constitution, the company’s public language continually revolves around responsibility and safety, and ethics itself becomes part of the product’s symbolic infrastructure. Anthropic 的聊天机器人 Claude 正是围绕与用户建立信任的理念构建的。该模型“响应”一套伦理宪法，公司的公开言论不断围绕责任和安全展开，伦理本身也成为了产品象征性基础设施的一部分。

The text of Pope Leo XIV’s encyclical Magnifica Humanitas. Photograph: Alberto Pizzoli/Getty Images 教皇利奥十四世的通谕《Magnifica Humanitas》文本。摄影：Alberto Pizzoli/Getty Images

“Magnificent” Humanity

“宏伟”的人性

The very title of the encyclical revolves around a specific tension: The human being is described as “magnificent,” yet at the same time capable of creating new forms of dehumanization. Seen from this perspective, artificial intelligence is not evil in itself. It is a mirror of the people who build it. 通谕的标题本身就围绕着一种特定的张力：人类被描述为“宏伟的”，但同时又有能力创造出新的非人化形式。从这个角度来看，人工智能本身并非邪恶。它是构建它的人类的一面镜子。

For this reason, the Vatican repeatedly warns about the risk of a new, “digital Babylon”: a society that reduces people, relationships, and even truth itself to data, performance, and efficiency. 因此，梵蒂冈反复警告一种新的“数字巴比伦”的风险：一个将人、关系甚至真理本身简化为数据、绩效和效率的社会。