The Hill favicon

Anthropic narrows AI safety policy pledge