Chances are, there would be.
Let’s assume for a moment that it is possible — or that it will be at some point — to write a perfect algorithm to govern our economies, our policy, our legal system, our traffic. There would have to be a data pipeline feeding the system enough information in real-time of course, for the algorithm to react to any situation by adapting its solutions in a timely fashion. The solutions provided by the algorithm will be optimal to all citizens, at least within certain geographic boundaries, but that’s fine — the system will adapt to cultural differences.
So. At what point in the preceding barrage of poppycock did you roll your eyes in skepticism?
That’s right. Not only is such “a perfect system” outside of our technological possibilities but well beyond our ability to decide on optimal criteria. In terms of deciding on policy for pretty much anything, we end up debating. There are different schools of thought on nearly any branch of formal knowledge: Freemanism vs. Keynesianism, Behaviorism vs. Cognitivism, Perfect vs. Imperfect Market Information, Strong vs. Weak Natural Law, Deontology vs. Utilitarianism, Welfare state vs. Society of proprietors, Big Rip vs. Big Crunch, Spinoza vs. Theism, OOP vs Functional, Dispensationalism vs. Premillenialism…
In short, human knowledge is imperfect — it’s all heuristics (experience acquired through limited sensory and rational abilities) or teaching (accumulated heuristics), and we tend to resist change. Even without thinking of the intervention of pre-existing interests, it’s easy to see that any selection process for optimality criteria to govern our hypothetical AI would be rife with selection bias.