Author: Nita Farahany Source: https://nitafarahany.substack.com/p/everyone-described-harm Published: February 9, 2026
"Everyone Described Harm" is the fourth installment in Nita Farahany's public account of her Advanced Topics in AI Law and Policy course, published February 9, 2026. It opens Week 4 on social media harms and platform liability. Farahany argues that the descriptive case for social media harm is now extensive and well-documented, and frames the remaining question as whether the legal system has a mechanism that can convert documented harm into liability at scale.
Summary of argument
The essay assembles a record of documented harm and then sets up the legal obstacle that record runs into: the liability shield in Section 230 of the Communications Decency Act. Farahany presents the distinction between a content theory of harm and an architecture theory of harm as the organizing frame for the rest of Week 4, and notes that the child-safety legislation covered later in the course implicitly takes a side in that debate.
The harm record
Farahany surveys evidence accumulated over several years:
- The Facebook Files (Haugen 2021). Internal Meta research disclosed by Frances Haugen showed Instagram knew its product worsened body-image issues for roughly 13% of teen-girl users, and that the company chose not to act on or disclose the findings.
- U.S. Surgeon General Advisory (2023). Vivek Murthy declared social media a youth mental-health concern and recommended labeling and design restrictions.
- Murthy v. Missouri (2024). A standing dismissal at the Supreme Court left the substantive jawboning question unresolved.
- MDL 3047 (Social Media MDL). Consolidated thousands of personal-injury and product-liability claims against Meta, TikTok, YouTube, and Snap.
- NM v. Meta jury verdict (March 2026). A New Mexico jury found Meta violated the state Unfair Practices Act on child sexual exploitation and mental-health concealment claims.
- LA bellwether (KGM trial, March 2026). A jury found YouTube and Meta negligent in product design and that the negligence was a substantial factor in the plaintiff's harm.
The Section 230 obstacle
Section 230(c)(1), codified at 47 U.S.C. § 230(c)(1), immunizes platforms from liability for content others created. The plaintiff strategy in MDL 3047 is to plead around Section 230 by attacking design — infinite scroll, autoplay, push notifications, and recommendation algorithms — rather than content. Courts have allowed some such claims to survive motions to dismiss, and Farahany describes the LA verdict as the first jury-stage validation of the design-based theory.
Architecture vs. content theory of harm
Farahany introduces a two-way distinction that organizes the remainder of Week 4:
- Content theory: harm comes from specific harmful posts. Section 230 fully bars liability, and the remedy is content moderation by the platform.
- Architecture theory: harm comes from design features such as engagement algorithms, variable-reward notification timing, and infinite scroll. Section 230 does not bar liability for the platform's own design choices.
She argues this distinction carries into legislation. Every child-safety bill covered in Week 9 — the Australia ban, the UK Children's Code, COPPA 2.0, and the CAADCA — implicitly takes a side. Bills that take the architecture side face fewer Section 230 problems but more First Amendment problems; bills that take the content side face fewer First Amendment problems but more Section 230 problems. Farahany argues there is no path that avoids both.
Provenance
The essay is the fourth post in Farahany's Substack series documenting her Advanced Topics in AI Law and Policy course (Class 4 of roughly 30). It is classified as a foundational source because it advances an original framing — the architecture-vs.-content distinction — that later course material builds on.
Relationships
- part-of: Nita Farahany Advanced Topics course (Class 4 of ~30)
- related: Garcia v. Character Technologies, Inc., AI Mental Health and Psychological Harm
- previous: Inside My Advanced Topics Class 3.2: Why Dark Patterns Work (Farahany, February 2026) next: Inside My Advanced Topics Class 4.2: The Shield — Section 230 (Farahany, February 2026)