I don't think it is feasible to be doing that on the server, unless you want games to cost even more. But also it's not that simple: suppose you don't want the client to be able to draw enemies through walls. But you do want the client to be able to play their footsteps through walls. So you need to send enough information for that - and now it's a matter of doing a small amount of computation to work out where an enemy is from the sounds being played, which can be much more effective than a human player's ears and brain.
I guess actually a simpler example is rendering enemies through smoke or fog - the game relies on them being hard but not impossible to see, and so the data needs to be there for a hack to potentially exploit.