A sudden evil thought just occured to me. Let's say you were a major corp, and interested in AI research, but worried about any AI you created getting loose into the grid and causing havoc.
Now say you owned prisons, and further say that you kept them off the grid.
And just for good measure, say you wanted to see how the AI would evolve, given slightly different parameters at 'birth.' and in slightly different situations, just to understand how stable or not your AI could be.
Now say another corp sends ShadowRunners in to figure what hush-hush operation is being run at those prisons ...
And this is why we don't let ourselves get caught by Mega-corps. Excuse me while I make sure my cyanide tooth is still viable.
On a slightly different notes, wouldn't loosing an extremely experimental AI into a prison's computer systems make actually running the system problematic. Would you let a hallucinogenic bi-polar sufferer run the prison? It could end up being one and the same if you're not careful, what happens if the AI suddenly develops a conscience and decides that what you're doing is unethical and decides the only ethical solution is kill the guards and free the prisoners? Even if such an attempt failed it could cause security breaches that would make a break out or infiltration possible that may not exist without the AI experiment being done.
Don't get me wrong, I love your idea but these are the pragmatic reasons I can think of not to do what you're suggesting as at the end of the day the prison has to remain operational and make money, no executive worth his job is going to risk that over AI research that may or may not pan out unless he is convinced somehow that the possible benefits outweigh the risk.
