“Anthropic's AI model, Claude, is trained to follow a set of principles known as a "constitution" to ensure its behavior is aligned with the company's safety goals.”