Unless the US government is hiding an unfriendly boxed AI and the Snowden disclosures have introduced a risk that this AI may unbox itself - no the downside can

Unless the US government is hiding an unfriendly boxed AI and the Snowden disclosures have introduced a risk that this AI may unbox itself - no the downside can not be anywhere near unlimited. How about we look at what we already know of the disclosures along with reasonable guesses of what remains, along with the probability that the remaining disclosures will ever become public knowledge, and judge him based on that.