Interesting stuff, following the saga of the recent WikiLeaks disclosure, they seem to have implemented some form of dead-man's switch on a future release of information.
http://cryptome.org/0002/wl-diary-mirror.htm
I would be curious to know if it's a fully automatic switch, maybe requiring a webpage to be visited, an email to be sent, etc. every x hours, or if it's human-based. Which would be more susceptible to attack, humans or code?
Reading up on AES256, it seems physically impossible to crack at this time. But with the decades spanning lead the military/intelligence community likely has on the private sector, it does make one wonder is the NSA, or other anonymous agencies, will be able to crack this file?
Other source:
http://www.wired.com/threatlevel/2010/07/wikileaks-insurance-file/?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+wired%2Findex+%28Wired%3A+Index+3+%28Top+Stories+2%29%29