[tomsmeding]

If you're talking about type systems, you typically have one per language: a language has a certain type system. A different language might have a different one. If different implementations of the same language have different type systems, they tend to be called different dialects, if not different languages altogether.

But as noted in a sibling comment by 'chc4, one might want different AI domains for the same language, either all at the same time, or in different compiler passes, or only if the optimiser is enabled, etc. While AI might look like a type system, and perhaps the algorithms are sort-of similar, at the very least the two are used differently.

EDIT: There is prior art on doing type systems and further analyses simultaneously; these are sometimes called "type and effect systems": https://en.wikipedia.org/wiki/Effect_system Contrary to what wikipedia seems to claim, such systems need not actually talk about side-effects only; one can also formulate provenance analysis in a programming language (e.g. which lambda can end up in this function-typed variable?) as a type-and-effect system.