Inside the Secretive Meetings Between Religious Leaders and a Leading AI Lab
Anthropic convened religious leaders, including Catholics, Jews, and Sikhs, to discuss how centuries of moral wisdom might inform the behavior of its AI models. Catholic moral theologian Charles Camosy, a professor at The Catholic University of America, participated in two meetings with Anthropic. Camosy said the discussions were held under the Chatham House Rule, limiting what he could share about individual participants’ comments. The meetings took place amid concerns about AI surpassing human intelligence and disagreements over whether AI could become conscious.
about 10 hours ago
Charles Camosy, a Catholic theologian who joined two meetings with Anthropic, described discussions involving prayer, theology and debate about how religious and philosophical traditions might inform AI systems. He said the meetings were governed by confidentiality rules intended to protect candid discussion, while stressing that he could not speak for the company as a whole or determine its motives.
Anthropic convened religious leaders—including Jews, Sikhs and Catholics—to discuss how “centuries of human moral wisdom” might help shape its AI models and products. Camosy said Anthropic led much of the presentations, which he found frustrating, but he also described the exchanges as good-faith debates.
Camosy recalled that the participants were deeply religious and that the meetings included prayer and substantive theological discussion. He said he raised the idea of the “Victory of the Cross” in the context of AI’s challenges, and that an Anthropic leader responded positively.
Camosy said he did not remember the precise terms of the nondisclosure agreements, but understood them as protecting participants as well as Anthropic. In his account, the agreements were not meant to hide the meetings’ existence; they were intended to encourage frank discussion and prevent unpublished materials from being shared prematurely.
He said he could not judge Anthropic’s broader intentions, which might be sincere or otherwise. The people he worked with directly struck him as sincere and not trying to pressure participants toward a particular position; he said they appeared to want disagreements not to obstruct areas of agreement.
Camosy regarded the idea of “forming” AI as potentially useful only as an analogy. He said AI systems are not persons and cannot themselves be more or less moral, while arguing that it is still desirable for them to behave honestly, carefully and in ways that respect human life.
He said he was uncertain how AI companies could reconcile input from different religious traditions, philosophers and their own commitments into a coherent ethic, though he suggested some overlapping consensus might be possible. On whether Anthropic’s leaders had been influenced by Pope Leo XIV’s position that AI cannot possess a soul, Camosy said he did not know and suspected the Church’s view would not substantially change their opinions about Claude’s moral status.
Camosy said cooperation with AI companies could still be important, particularly in addressing catastrophic and existential risks. He also described engagement beyond Anthropic: he helped plan a meeting involving U.S. bishops’ administrative leaders and representatives from frontier AI labs, and said Google DeepMind and OpenAI had invited him to speak about Magnifica Humanitas.
Camosy said his view of advanced AI had shifted from seeing large language models mainly as sophisticated next-word or next-pixel predictors to regarding their operation as more complex and mysterious. He maintained that the models are not conscious or persons, but said uncertainty about what is being built and how it functions should itself be a serious concern.
Catholic ethics must orient artificial intelligence toward human dignity
Catholic ethics should shape AI from its design through its everyday use, making human dignity—not efficiency, profit, or technical capability—the decisive measure. The Church’s 2025 note Antiqua et Nova teaches that an AI system’s goals, methods, and underlying vision must respect every person’s dignity and serve the common good.
This dignity does not depend on intelligence, productivity, or ability. Because every person has inherent worth as one created in God’s image, AI should not reduce people to data, outputs, or their usefulness to society.
AI can assist human decisions, but it cannot bear moral responsibility: people design systems, choose their purposes, and remain accountable for their consequences. Accordingly, developers, owners, operators, regulators, and users all have a responsibility to ensure that AI supports human freedom and sound judgment rather than displacing them.
The practical test is therefore not simply whether AI works, but whether its ends and means promote human flourishing, justice, fraternity, and the common good—and whether its harms to people or communities are avoided. That judgment calls for prudence and responsibility at every level of society.