Every commercial language model, recommendation system, and content ranker is embedded in surveillance capitalism. The training data comes from surveillance; the economic justification for investment comes from prediction and influence; the alignment question becomes, Zuboff argues, inseparable from the question of whose interests these systems serve.
Zuboff's analysis has become increasingly applicable since 2022, as generative-AI systems have entered the same commercial context they critique. Training data for frontier models is harvested from the same substrate — user-generated content, behavioral traces, scraped records — that Zuboff's earlier analysis centered