A look at the UK's AI Safety Institute, whose researchers probe AI models for safety gaps, as its work becomes a blueprint for other governments' AI policies (New York Times)