Is there a proof that states the possibility or impossibility of an AI system to acquire more sophisticated capabilities (in terms of generic cleverness) than its own creator?
Regarding Artificial General Intelligence, which does not currently exist and is still highly theoretical, this cannot be determined at this time.
What I would say is that "strong narrow AI" has already proven the ability to become "smarter" than it's creators in specific tasks. (See Alphago, etc.)
Under the idea that some form of AGI might come out of an algorithm comprised of an ever expanding set of strong narrow AIs, it would logically follow that such an automata could become smarter than it's creators in any given task.
The loaded term in your question is "generic cleverness." There's no such thing. What is "smart" is only smart relative to a criterion. Provide a complete criterion and we can talk rationally about "levels of sophistication." Until then, there is no measure against which "involuntarily acquired" capabilities can be regarded as more or less "smart."
If you look at the work of Howard Gardner and his theory of multiple intelligences, you will see that the term of "intelligence" respectively "generic cleverness" is much more diverse and not entirely clarified. Without a entire notion of it, a proof is forlorn.