Lets, be honest, it's not like academics is the only place this happens. CS labs are essentially small companies (1 CEO, 2-3 senior members, and 5-10 junior members). The only difference is that the company produces papers, producing usable code that you support is actually very hard and takes away from the work that actually matters. I'm actually much less worried about being scooped by releasing my code, and much more worried about having to answer emails asking me how to get the hacked together code to run on their machine or their dataset.