As long as you have a communications channel to the AI it would not be secure, since you are not a secure system and could be compromised by a sufficiently intelligent agent.
Intelligence is no help if you need to open a safe that only gets opened by one of the 10^10 possible combinations. You also need enough information about the correct combination to have any chance of guessing it. Humans likely have different compromising combinations, if any, so you'd also need to know a lot about a specific person, or even about their state of mind at the moment, the knowledge of human psychology in general might not be enough.
(But apparently what would look to a human like almost no information about the correct combination might be more than enough to a sufficiently clever AI, so it's unsafe, but it's not magically unsafe.)
If you had a program that might or might not be on a track to self-improve and initiate an Intelligence explosion you'd better be sure enough that it would remain friendly to, at the very least, give it a robot body, a scalpel, and stand with your throat exposed before it.
Surrounding it with a sandboxed environment couldn't be guaranteed to add any meaningful amount of security. Maybe the few bits of information you provide through your communications channel would be enough for this particular agent to reverse-engineer your psychology and find that corre...
In the early 1980s Douglas Lenat wrote EURISKO, a program Eliezer called "[maybe] the most sophisticated self-improving AI ever built". The program reportedly had some high-profile successes in various domains, like becoming world champion at a certain wargame or designing good integrated circuits.
Despite requests Lenat never released the source code. You can download an introductory paper: "Why AM and EURISKO appear to work" [PDF]. Honestly, reading it leaves a programmer still mystified about the internal workings of the AI: for example, what does the main loop look like? Researchers supposedly answered such questions in a more detailed publication, "EURISKO: A program that learns new heuristics and domain concepts." Artificial Intelligence (21): pp. 61-98. I couldn't find that paper available for download anywhere, and being in Russia I found it quite tricky to get a paper version. Maybe you Americans will have better luck with your local library? And to the best of my knowledge no one ever succeeded in (or even seriously tried) confirming Lenat's EURISKO results.
Today in 2009 this state of affairs looks laughable. A 30-year-old pivotal breakthrough in a large and important field... that never even got reproduced. What if it was a gigantic case of Clever Hans? How do you know? You're supposed to be a scientist, little one.
So my proposal to the LessWrong community: let's reimplement EURISKO!
We have some competent programmers here, don't we? We have open source tools and languages that weren't around in 1980. We can build an open source implementation available for all to play. In my book this counts as solid progress in the AI field.
Hell, I'd do it on my own if I had the goddamn paper.
Update: RichardKennaway has put Lenat's detailed papers up online, see the comments.