This issue was addressed in the white paper. An investor collective successful in attempting to game rewards for themselves would cost them more in terms of capital loss than they would gain from payouts. If you assume they gamed it to earn 100% of all rewards, the price would fall by at least 10% and they would be locked into vesting STEEM unable to escape.
There are enough good people around to downvote that kind of attack that it is unlikely to ever be an issue.
Bad wording on my part. I didn't mean it as an attack vector. The active invested parties are currently deciding what people see when they enter Steem. They upvote based partly on what they think the other Steem investors will vote for (short term), what they think is objectively good content (long term), and what their own personal bias/tastes are (irrationality). The short term bias is obvious, and it might even help Steem at this stage to focus on internal issues, but eventually what investors are voting for will have to reflect what steem users are interested in, and it's not clear exactly how this will happen. What investors in Steem think is objectively good content is also something that will reflect their own bias, e.g. they might think content is "dumb" when in fact it may be suited for an ideal steem target group. Over time the people holding Steem will be the ones making the most accurate judgements about upvotes for the current investor group, not necessarily for the ideal target group. The plans of giving everyone a vote is probably going to help a lot with this problem, there may be many other fruitful ways to deal with it as well.