Since chatbots went mainstream in 2022, a security flaw known as prompt injection has plagued artificial intelligence developers. The problem is simple: language models like ChatGPT can’t…