<< 2 points:
1.) Having certain restrictions might be a spur to the AI to find alternative harmelss solutions to a problem;>>
I would say that restrictions might prevent an AI from implementing obvious harmful solutions to a problem, though as well as channelling them to harmless solutions it may also send them toward less obvious harmful ones. Nothing is perfect (yet).
<< 2.) The freedom of the AI to develop must be balanced with the possible effects on humans. Speaking selfishly, I'd plump for the latter party.
I'd say that the freedom of any AI must be balanced against the possible effects on any sentient, human or AI, just as a human's freedom must be balanced against the same possibility of causing harm. I just hope that with designed intelligences we can do a better job of instilling a respect for others than we tend to do now.