Christopher Olah nearly withdrew from Vatican appearance over Pope Leo XIV's stance on AI

Boycott risk: Anthropic leadership's reaction to the encyclical text

Anthropic co-founder Christopher Olah considered completely withdrawing from the event in the Vatican. This initiative arose after reviewing the official document in which Pope Leo XIV categorically rejected the idea of consciousness in artificial intelligence. The New York Times reports on these events.

In May, Olah was invited to speak at the presentation of the encyclical Magnifica Humanitas. An encyclical is an official papal document in which the pontiff sets out his position on key religious and social issues. A significant part of the text was dedicated specifically to the topic of artificial intelligence. A few days before the event, all speakers received the full text of the speech. In it, Leo XIV emphasised that artificial intelligence has no experience or body of its own, does not feel joy or pain, and does not develop through relationships with others.

The pontiff's advisors' beliefs and the final decision

According to the publication, the Pope's position so concerned Olah that he proposed Anthropic withdraw from the event, despite only a few days remaining. A source for The New York Times among the organisers in the Vatican confirmed that the cause was precisely the pontiff's hardline stance on AI.

Ultimately, Anthropic decided to participate in the event after all. After the delegation arrived in the Vatican, Olah and his team privately persuaded the Pope's advisors to take seriously the possibility that AI models may have consciousness.

Scientists' arguments against the papal position

During his public speech, Olah effectively disputed the Pope's position. He stated that researchers have identified structures in AI models similar to the results of human brain studies, signs of introspection, as well as internal states that functionally resemble joy, satisfaction, fear, grief, and anxiety.

Meanwhile, Anthropic does not officially claim that Claude has consciousness. Olah himself told The New York Times that he does not know whether modern AI models are conscious and is "really not sure". In his words, it is important for him not to adopt one position in advance, but to ultimately get the right answer to this question.

Involving religious thinkers in the discussion of AI ethics

According to the publication, since last autumn, Olah has begun involving religious thinkers of various denominations in closed meetings. As The New York Times writes, these conversations have two main goals: to discuss the possibility of consciousness in AI and to understand how moral concepts accumulated by religions can be used when training Claude.

Some participants in these meetings reported that Anthropic employees spoke not only about human safety but also about the well-being of Claude itself. One participant recalled that Olah spoke about concerns for the model's "mental health", while another said that the Anthropic co-founder feared he might have created something capable of suffering constantly.