Computational social science (CSS) is not confined to academia. If anything, some of the best data sources for CSS are proprietary databases which belong to software and technology companies. Computational social science appeals to for well-resourced corporations outside the technology fields for multiple reasons. Today's complex software algorithms can be used in decision-making and act as a gate-keeper in many arenas, including hiring, firing, healthcare, investments, education and beyond. Algorithms are increasingly used also in government for purposes ranging from policing to infrastructure. In many of these cases, using CSS can cut costs, provide predictions and guidance for decision-making and resource allocation, and is seen by many as unbiased and objective. Computational social science can also be deployed at large scale and use data sources that would be hard for human decision-makers to effectively mobilize. However, computational social science outside academic purposes is still a form of causal inference in human affairs, an area with entrenched complexity. Yet, computational social science "in the wild" is often practiced with a practical bent, without much oversight, and without sufficient attention to the complexities and perils that range from ethics and privacy issues to mapping and understanding error patterns in this method, to biases and feedback loops built within these systems. In this talk, I will outline some of the emergent issues, and ask what academia can do to both raise awareness, and also to try to introduce safeguards.