They're creating an infrastructure of manipulation because of the collective impact of the industry within the current legal context of international privacy protections. Any individual project or feature can look trivial or harmless, like manipulating the order of articles in a feed, but the systems used to do it are mechanisms implicitly susceptible to bad incentives.
They're collectively pressing buttons and breaking things in the global psychology, influencing the zeitgeist in a way that no one person or company intends. Obama's campaign was the first major successful political use of that infrastructure and there's an argument to be made that Trump's election was enabled by it. Cambridge Anallytica exposed some of the controls and showed a portion of what can be done, but we haven't seen the worst yet. International troll farms, the recent infiltration of Christian and BLM Twitter and reddit, and so on.
When people are profiled by advanced algorithms, and access to people given based on those profiles, that access can be made completely transparent using deanonymizing methods through data from other sources. Any and every platform, forum, and community becomes exposed to hostile, unintended influence.
We need outsized fines that protect private data as if it were sacred such that a trillion dollar company would fear even a single violation. Such that violating privacy would result in individuals going bankrupt or prison time. Such that physically bombing a foreign datacenter would be seen as reasonable recourse if they were uncooperative.
There are things that no government or private organization should have absent the informed consent of users because of the implicit danger to society's basic functioning.