An AI researcher standing next to a pope is not a sentence anyone had on their bingo card. But on May 25, 2026, that’s exactly what happened when Anthropic co-founder Chris Olah delivered an address at the Vatican during the unveiling of Pope Leo XIV’s new encyclical on artificial intelligence and human dignity.
Olah, who leads the mechanistic interpretability team at Anthropic, used the occasion to argue that AI models are far more complex than most people realize. They may even harbor internal states resembling emotions, he said, which raises ethical questions the tech industry alone is not equipped to answer.
What Olah actually said, and why it matters
Olah called for independent guidance from entities outside the tech ecosystem, specifically naming religious institutions, governments, and civil society organizations as necessary counterweights to the industry’s internal momentum. He cited conflicting pressures on AI labs — commercial incentives, competitive dynamics, and the sheer speed of development — as creating blind spots that make those labs unreliable as sole stewards of the technology they’re building.
He also warned that AI has the potential to displace human labor “at very large scale.”
