Elm is different to ML in that it is explicitly designed to be a pure language. Purity is important for the Elm architecture and allowing arbitrary side effects in your apps would break everything.
That said, you could add effects and a JavaScript FFI using monads or algebraic effects. You could also get rid of a lot of boilerplate in the Elm architecture using existential types. It's a safe bet that the Elm developers know this. But all of these extensions would make the language less approachable. I think one of the reasons why Elm is doing so well is because it's not Haskell. :)
That said, you could add effects and a JavaScript FFI using monads or algebraic effects. You could also get rid of a lot of boilerplate in the Elm architecture using existential types. It's a safe bet that the Elm developers know this. But all of these extensions would make the language less approachable. I think one of the reasons why Elm is doing so well is because it's not Haskell. :)