AI and surveillance capitalism are undermining democracy (thebulletin.org)

On March 6, 2025 Axios revealed the State Department rolled out “Catch and Revoke,” an AI-assisted social‑media surveillance program intended to scan “tens of thousands” of student‑visa holders’ online footprints for alleged post‑Hamas‑attack sympathies. The announcement crystallizes a broader, systemic problem: AI’s power to predict behavior is being braided into surveillance capitalism and government practice, blurring private and public life and enabling quick weaponization of data for enforcement, deportation, or political control. High‑risk examples cited include period‑tracker data turned into legal evidence, Ring and Echo feeds subpoenaed for policing, and precedent‑setting privacy lawsuits against Meta and Flo. Technically, the story highlights how modern ML—through high‑dimensional vector embeddings and a zoo of methods (supervised, unsupervised, online, reinforcement learning)—optimizes pattern extraction and group‑level prediction rather than true individual understanding. That “smoothing” turns people into class members whose average behavior drives decisions (loans, admissions, border enforcement), creating a one‑way surveillance ratchet: more AI needs more data, which is monetized and redeployed. The result is biased, decontextualized inferences, chilling effects on free expression, and concentrated power across public and private actors. The piece argues these risks aren’t solved by better models alone but require restrictive, well‑designed legal and technical guardrails to protect democratic norms.