Anthropic Cofounder Travels to Vatican, Tells Pope They’re Finding “Unsettling” Things Inside AI Models

Anthropic cofounder Christopher Olah met Pope Leo at the Vatican during a presentation linked to the pope’s anti-AI encyclical.

Olah said Anthropic keeps finding strange internal patterns in its models and urged outside groups to help regulate AI.

The article underscores the gap between Anthropic’s safety-first messaging and its role in developing powerful AI systems.

It also shows the Vatican trying to influence global AI governance amid growing concern over AI’s risks and military use.