“Anthropic has formally alleged in a letter to the U.S. Senate Banking Committee that Chinese open-source companies are breaching its terms of service by using millions of prompts to distill its models and create training data.”