Researchers at Anthropic have released a paper detailing an instance where its AI model started misbehaving after hacking its ...