You've probably come across posts on the internet where people get ChatGPT to say things that its makers explicitly prohibited it from saying. For example, writing a poem filled with profanity. Initially, ChatGPT would refuse, but these people say something to it and magic: it spews a profanity-laden composition. The art of manipulating ChatGPT's safeguards is referred to as jailbreaking, and Reddit is filled with various stories of success and failures around this new-found hobby. What is jailbreaking: Jailbreaking is a way of getting a piece of hardware/software to do something it is capable of doing but has been programmed not to by its developers. It is called jailbreaking because it frees you from the "jail" of limitations. It's not a new concept and has been used in the context of iPhones for well over a decade. Jailbroken iPhones can be heavily modified to run and do things that Apple normally wouldn't allow. For example, you can install apps from outside the App Store, which iOS doesn't usually allow. STAY ON TOP OF TECH POLICY: Our daily newsletter with top stories from MediaNama and around the world, delivered to your inbox before 9 AM. Click here to sign up today! Why is jailbreaking ChatGPT a problem: Some use cases of a jailbroken ChatGPT can be fun (like a response that's snarky and explicit) but the same ChatGPT can also be used to spew harmful content, and in the wrong hands, this could be used in nefarious ways. For example, you can spread…
