You seem to be suggesting that the point of decentralisation is to avoid legislation and regulation. While this might be true for Silk Road and other criminal transactions, that does not seem to me to be a valid argument for running a prediction market on blockchain software.
Blockchain as a technology has a valid use-case where you cannot trust a third party to manage a centralised database and where you need an immutable history of events. I can just about buy into putting transactions for real-world things (eg property) onto a blockchain - after all its a lot easier to prove you own something in the case of a dispute if you can show on a distributed and trusted event log that you were the last person to purchase it than it is to dig out some old documentation and prove its not forged.
There is an argument that applies to prediction markets, which is that you might not trust the market to actually pay out your winning contracts (though believing in this relies on a distrust of the regulators and court system). Augur is not actually solving this problem though, as neither oracle contracts nor the market owner defining the settlement conditions are trustworthy, and when you have user-defined markets such as "Does God Exist?" there is huge scope for arguing about the true outcome. At least with the centralised case you have an organisation with a reputation to maintain, that is therefore motivated to police and administrate the markets properly.
You might also argue for long-running markets that the company running the market might collapse and be unable to pay out, but in the case of crypto there is simply a parallel that your currency could simply collapse and be worthless by the time your contract paid out. Given the volatility of most ETH and REP, your winnings are likely to either be dwarfed by gains in the currency or destroyed by a collapse in the currency.
Prediction markets that trade using real money are gambling, and hence are controlled by various regulators. In addition, certain markets can also catch the attention of other regulators, for example if a market involves speculation on financial instruments or commodities. There are certain ways to avoid this, such as restructuring the incentives; for example, a market using credits instead of real money and run as a competition where the most successful players receive monetary prizes does not constitute gambling. Obviously in this case the players are not risking their own money, and this relates back to my previous points about subsidising prediction markets in exchange for valuable information. The regulation issues should be dealt with in an ethical way by developing the technology and working with regulators and not by avoiding the issue via decentralised blockchains.