I think it might be do to the copied string ending in \n or something like this. It shouldn't be hard for Wikidata to accept the copy pasted string.

I'm not sure about this specific copy-paste, but I think it's a frequent problem when strings start or end with odd characters or even include CR in the middle.

It happens a lot when trying to create statements for article titles. Oddly, I think the same text for labels tends to work.

Instead of rejecting it completely, it would be helpful if a cleaned version was offered for saving instead.

Please distinguish between

a) normalization applied by the API, when receiving a JASON structure representing a sting value.
b) normalization applied by the API when parsing user input, and returning a string value
c) normalization applied by the JavaScript widget before sending user input to the API

I believe we should do (b), but not (a). We can also do (c), but with (b) in place, this seems redundant.

The reason I believe we should not do (a) is that we are not dealing with user input then, but with a DataValue object. DataValue objects are either valid or they are not, we should not guess intent. Clients that supply DataValues should take care to apply any normalization they desire.

I agree that (b) is likely the way to go.