You put those safety comments in two places: when you declare an API as being unsafe, and when you use an unsafe API. Not coincidentally, those are the two plac

You put those safety comments in two places: when you declare an API as being unsafe, and when you use an unsafe API. Not coincidentally, those are the two places in the language that force you to use the “unsafe” keyword. If you declare a safe API that uses unsafe under the hood, it’s on you to ensure that it’s safe to call under any situation, so that callers can call it without worrying that their program is suddenly unsafe. If you can’t guarantee that your API is safe under all circumstances, you need to declare it unsafe and make it the caller’s job to use it safely.