Anthropic flew scholars from Catholic, Jewish, Sikh, and evangelical traditions to its San Francisco headquarters to develop a "moral compass" for its Claude AI model. These guests, including Vedanta monk Swami Sarvapriyananda, signed nondisclosure agreements for closed-door sessions with mental health professionals and one of the company's founders to assist in training the system.

The company's primary goal during the encounters was to persuade these leaders that its AI possesses a soul and moral standing, according to researcher David DeCosimo. The New York Times first reported that Anthropic had been quietly convening global religious thinkers to advise on AI safety, while critics have described the secrecy of the meetings as disturbing.