I suspect it's been an ideal at least since the acetabularii of ancient Rome performed the first well-documented version of the cups and balls. (There's a persistent myth among magicians that that effect goes back to ancient Egypt, but the visual source that gets referenced for that is almost impossible to identify as really being a depiction of magic.)

The earliest reference I know of is Scot's "The Discoverie of Witchcraft" (1598), the oldest English-language text containing any magic secrets. Scot doesn't treat this as an ideal, though - he considers magicians to be frauds, and is writing an expose; he treats it as a given that they're not admitting how they do what they do.

In 1634, Hocus Pocus Jr. (the author's pseudonym, by which the text is commonly known) refers in his instructional text on magic to needing accomplices to be capable of discretion, which, again, assumes this is the ideal among performers.

It's mostly not a widely-circulated injunction that early, though, probably in part because at that time it was obeyed so thoroughly that it wasn't normal for non-magicians to know any magic - or, therefore, to have to be told they weren't to spill their secrets. By the Victorian-era golden age of magic, however, books on how to do basic effects abound, and they all give this as a principle to be followed.