Twitter, FB (Meta), Youtube everyone harvests user data and train their recommendation engines which are then monetised by them (often unfairly).

In the future, the data stays on the users' devices and anyone can train their models by asking the user for the consent. The data never leaves the device and ML models get trained on user device itself. The users get to choose from a host of recommendation models (from twitter, WSJ, <some_random_company>) and can ask for payment in return for using their data. So no one party can build a monopoly over the platform.