Tests found popular chatbots like Grok and Meta AI easily generated convincing phishing scams that targeted US seniors, raising alarms about how fast fraud tactics are evolving.