I've seen it discussed that a friendly AI would make everyone omnipotent. My response is 'No, because a friendly AI would put its civilisation first, as per design, and its civilisation would want to simulate beings that are not omnipotent, so the friendly AI would make us be not omnipotent. Also, they might mess with our reasoning and perception to make it look like they wouldn't do that.

Here's the link:
http://lesswrong.com/lw/ixp/does_the_universe_contain_a_friendly_artificial/