The immediate and obvious problems stem from self-satisfying actors. "Oh, 10,000:1 on blowing up Country Y's embassy in Country Z this week! Lets go make some money boys."
If a terrorist plans to blow up an embassy and places bets on this action, he's directly transmitting his plans to the public. The price of "embassy bombing" skyrockets, security agencies are notified, and his odds of success go way down. (His odds of surviving the attack and collecting his money also aren't that high anyway.)
Now consider the possibility that the terrorist leaks information to his supporters/criminal underworld/etc - e.g. the gangsters selling him semtex. These guys would never betray him, but on the other hand they see some free money sitting on the table. They place bets on an attack, the price goes up, and again agencies are notified. That's a good thing - a truly anonymous way for criminals to probabilistically inform on each other.
When smart people like Robin Hanson (who was in part behind this proposal) and various CIA types get behind a proposal, you should always think twice about the "immediate and obvious problem". The folks proposing this aren't stupid, so they probably thought about it too.