I don't understand why I was downvoted for asking a question. I just meant to imply that there must exist some middle infrastructure, and I do not believe that it is a trivial problem to solve.
Source: I studied computer security systems that attempted to automate the flow of information to maintain security level thresholds on information types. When code 'touches' other code, meaning that it uses code with information at one security level to reason about code with information at another security level, unless you are particularly careful in the structure and organization of the code, information can leak out, structures can be manipulated, rules can be broken.
I am not saying scientists are unethical, I am saying they are human and prone to emotional reaction occasionally before objectivity (unless they are perfect). Combining information inference, intelligence, and an understanding of the flaws of an automated system; they are capable of 'accidentally' tweaking those things in their favor. It's not in the interest of science when this happens, and it requires a substantial amount of real intelligence to moderate. It's basically 'accidental bias'.