How Meta (Facebook) does it
Meta's business model is named in their own SEC filings. From Meta's 10-K disclosures: substantially all of Meta's revenue comes from advertising. Specific 2024 figures: total revenue $164.5 billion, advertising approximately $160 billion (~98%). The remainder is Reality Labs hardware. The advertising product is targeting precision: advertisers pay Meta to reach specific users defined by inferred attributes — age, location, interests, recent purchases, social graph connections, content engagement patterns, and behavioral signals collected across Facebook, Instagram, WhatsApp, and Meta's tracking pixels embedded across the broader web.
Meta's lawyers distinguish between "selling user data" (which they technically do not, in the strict sense) and "selling targeting that uses user data" (which is the entire $160B business). The Constitution treats both as the same act: monetizing the user as raw material. The Cambridge Analytica scandal — for which Meta paid a $5 billion FTC settlement in 2019, the largest privacy penalty in US history at the time — was not a one-off. It was the visible portion of an architecture where user data is the asset being mined and refined into targeting precision. Frances Haugen's 2021 disclosure (cited in Refusal 1) showed Meta's own internal researchers documenting harm to users and the company suppressing the findings to protect the engagement metrics that drive ad revenue. The architecture extends across the industry: Google ($240B+ in 2024 ad revenue), TikTok/ByteDance, Twitter/X, Snap, Reddit — every "free" social platform funded by advertising-on-targeting.
The harm reaches further than the obvious one. When user data is the asset, every product decision has a hidden vector: what increases engagement also increases inferred-data fidelity, which increases targeting precision, which increases ad revenue. The user becomes the optimization target. Meta's internal research showed harm; the architecture made the harm profitable; the architecture kept going.