Researchers at Anthropic, the company behind the Claude AI assistant, have developed an approach they believe provides a practical, scalable method to make it harder for malicious actors to jailbreak ...