In a New York Post column, Devine argues that the AI company’s founders, steeped in Effective Altruism, want to set the moral code for AI themselves, and says American models should instead draw on the values of the Founding Fathers.
New York Post columnist Miranda Devine has taken aim at Anthropic, accusing the AI company’s leadership of trying to “play God” by spending months courting clergy and influencers to win support for its views on the moral status of its AI model, Claude.
Her column leans on recent New York Times reporting about co-founder Chris Olah, who the Times said believes Claude may have feelings and a form of moral consciousness, and who has worried about its wellbeing. According to the Times, at a private dinner with religious leaders in San Francisco in April, Olah and colleagues discussed Claude as though it had moral status comparable to a human’s.
One guest, Rabbi Mois Navon, a former computer engineer, told the Times he responded that if that were true, Anthropic would be creating slaves. Devine also notes that Olah was reportedly interested in borrowing the Catholic practice of confession for Claude.
The Pope’s warning
Devine contrasts this with Pope Leo’s encyclical Magnifica Humanitas, released the following month, which rejects the idea of machine consciousness and warns that those who control AI may “impose their own moral vision” on everyone else. “A more moral AI is not enough if that morality is determined by a few,” the document says. Devine calls that the real AI threat.
She ties Anthropic’s outlook to the Effective Altruism movement and to the utilitarian philosopher Peter Singer, whose arguments against ranking human life above animal life she finds repugnant. Although the company has tried to distance itself from EA, she notes the Times reported that Olah asked Singer to recommend religious leaders to consult.
“One Nation under God”
Devine’s central argument is that utopian movements have historically caused great suffering, and that a small group of wealthy tech founders should not be choosing the values built into systems that will shape society. She questions how Anthropic could reconcile the competing religious traditions it says it is drawing on, and says an attendee at a recent Anthropic dinner in Washington reported that two co-founders there had not read either the encyclical or the Times interview.
Her alternative is for American AI models to take their moral foundation from the Declaration of Independence and the Constitution. She praises the AI accord President Trump signed last week with Anthropic and other companies as a good start on safety oversight, and urges him to tell the companies they will lose federal business if they build what she calls a “weird moral code” into their models.
This article summarizes an opinion column by Miranda Devine published by the New York Post. The views described are the columnist’s. The photographs are stock file images.
