I'm wondering if there's a maxim that is morally permissible that cannot be universalized, or at least done so consistently under Kantian ethics and his formula of universal law. In Examples of universalizable maxims in Kantian ethics the maxims remain valid when applied universally to everyone.
"Act only according to that maxim by which you can at the same time will that it should become a universal law." The ambiguity is in what you wish to call a "universal law." Is it simply any universally quantified rule consistent with the specific case, like ∀x ∀y ~Steal(x, y) ? Or are there more specific conditions?
If it's simply any universally quantified rule, then what prevents ∀x ∀y DaveHaskins(x) ∨ ~Steal(x, y) ? This seems like a universalization attractive to a career thief named Dave Haskins, and consistent with the maxim he lives by. In what way is ∀x ∀y DaveHaskins(x) ∨ ~Steal(x, y) less "universal" than ∀x ∀y ~Steal(x, y) ?
"Oh no, but it names a person!" you might object. Well, it turns out Dave Haskins happens to drive a car with license plate 25306. So, what prevents him from universalizing in the following way: ∀x ∀y CarLicensePlate(x, 25306) ∨ ~Steal(x, y). Perhaps there are others with the same license plate in other states, so it doesn't name only him, but Dave would still be happy with this universal law.
"But you're adding extra conditions!" you might object. Well, what is this predicate Steal(x, y) but a whole box of conditions saying when an act is stealing or when it is not? It is only possible to steal what is someone else's property, that they do not give willingly or in trade. So to give the same meaning as ∀x ∀y ~Steal(x, y), we may say, ∀x ∀y ∀z ~ (OwnsProperty(z, y) ∧ Takes(x, y) ∧ ~GaveWillingly(z, x, y) ∧ ~TradedAway(z, x, y) ). And really this is not a complete picture; to fully specify the meaning of the word "Steal" we would have to add many more conditions.
So, if adding extra conditions is not a valid objection to Dave's selfish universalization, and neither is the fact that Dave names himself, on what grounds can you object to it?
I'd like to further note that examples like Dave's are common in real life; people invent ostensibly "universal" morality to justify whatever they happen to benefit from. Social Darwinism comes to mind, or "manifest destiny," or the mandate of Heaven for Chinese emperors, or the justifications for keeping slaves made up by slaveowners.
Peter Singer's idea that moral progress involves expanding the circle of moral concern, involves development. And suggests other species can become, are becoming moral agents. Such a developing ethic cannot be universalised prescriptively for all times and places, I think. It is an ethic that changes.
Religious ethics often involve a higher standard that can be achieved, such as by saints (or monks or anchorites), bodhisattvas (or the sangha), or the Sikh kalsa. They represent aspirational moral codes. Like the demands on volunteer soldiers, or secret service agents, there are demands on behaviour which cannot be demanded of everyone. We hold police & politicians 'to a higher standard', implicitly making it non-universalisable - we can note that preventing people who's businesses have ever failed from starting new ones, is a massive economic drag, to understand why. Sometimes it's appropriate to demand not a single failure. But also we should not make people' s pasts entirely define them.