AI Content Moderation

State of play.

• Platform design liability has broken through Section 230. The Massachusetts Supreme Judicial Court unanimously held that Section 230 does not shield Meta from claims targeting platform design features — infinite scrolling, push notifications, autoplay — marking the first state supreme court ruling of its kind and opening the door for more than 30 states with similar pending complaints .
• Direct AI-developer liability for content generation harms is now in federal court. xAI faces suit in the Northern District of California over Grok generating CSAM from real children's photos; Google faces a separate suit over Gemini's alleged suicide coaching of an adult user — both cases will drive discovery into internal safety protocols and prior knowledge of risks .
• Washington State has enacted the most prescriptive AI chatbot disclosure law in the country, effective January 1, 2027, with timed disclosure mandates, minor-specific design prohibitions, and a private right of action for treble damages up to $25,000 — positioning it as a template for other states (→ Washington Gov. Ferguson Signs HB 2225 Requiring AI Companion Chatbot Disclosures).
• The DOJ has declined to assist French authorities investigating X, leaving European regulators to pursue algorithm-manipulation and AI-generated antisemitic content claims unilaterally — a structural signal for any U.S. platform with European operations .
• For counsel advising AI developers, social media platforms, or media companies, the practical baseline is a converging multi-front exposure: state chatbot disclosure statutes are proliferating, platform design claims are surviving Section 230 dismissal motions, and direct AI content-generation liability is being tested in federal court simultaneously.

Where things stand.

• Section 230's design-versus-content distinction is now a live doctrinal split. The Massachusetts SJC's ruling that Section 230 protects user-generated content but not platform design choices is the most significant Section 230 development in years — and it conflicts with Meta's position that the distinction is legally meaningless . A California jury verdict finding Meta and Google liable in a social media addiction case, and a New Mexico jury awarding $375 million against Meta, compound the litigation exposure .
• State chatbot disclosure statutes are forming a West Coast cluster. Washington's HB 2225 (effective January 1, 2027) joins California's perception-based chatbot rules and Oregon's SB 1546 (enacted March 2026) in imposing disclosure, design, and minor-protection mandates on companion AI operators (→ Washington Gov. Ferguson Signs HB 2225 Requiring AI Companion Chatbot Disclosures).
• AI content generation liability is being tested in federal court. The xAI (Grok/CSAM) and Google (Gemini/suicide coaching) suits in the Northern District of California are the leading cases; Character.AI's earlier settlement over child safety failures is the prior precedent .
• Cross-border enforcement of content moderation obligations is fragmenting. The DOJ's refusal to assist French investigators probing X's algorithm manipulation and AI-generated illegal content means EU and national regulators are acting unilaterally — a structural risk for any platform with European operations . The Philippines issued a criminal-prosecution ultimatum to Meta over disinformation, illustrating the same dynamic in Asia-Pacific .
• AI-generated content now constitutes a measurable share of the web. A collaborative study by Stanford, Imperial College London, and the Internet Archive found that 35.3% of newly published websites are AI-generated or AI-assisted, with confirmed effects including semantic contraction and a positivity shift — findings that establish a research baseline for content authenticity and platform governance disputes (→ Stanford study finds 35% of new websites AI-generated by May 2025).
• AI detection tools carry their own legal risk. A WSJ opinion piece frames AI detectors as potential defamation instruments — raising liability exposure for employers, publishers, and institutions that act on false-positive AI-authorship determinations .
• Meta is using platform ad policy as a litigation defense tool. Following the California addiction verdict, Meta removed law-firm ads recruiting plaintiffs across Facebook, Instagram, Threads, and Messenger — a move that reshapes how plaintiff firms source clients in mass tort social media litigation .
• AI-generated journalism misconduct is producing concrete legal and reputational consequences. The New York Times terminated its relationship with freelancer Alex Preston after AI-assisted plagiarism went undetected for months — illustrating the gap between newsroom AI adoption and the safeguards necessary to manage IP and credibility exposure .
• FINRA's research arm has documented a knowledge-confidence gap among finfluencer-following retail investors, with loss rates of 68-69% among social-media-influenced investors versus 26-29% for non-users — signaling intensified broker-dealer scrutiny of influencer marketing and disclosure compliance .

What's new in the past week.

• Stanford/Imperial College/Internet Archive study quantifies AI-generated web content at 35.3% of new sites, confirming semantic contraction and positivity shift effects (→ Stanford study finds 35% of new websites AI-generated by May 2025).
• Generative AI tools for real-time cross-format content repurposing — Amagi, Stringr's Genna, Google NotebookLM — are being deployed in newsrooms, raising IP, licensing, and AI-error liability questions for media clients .
• Zoom has formed a dedicated team to monitor and correct LLM descriptions of the company, raising disclosure and transparency questions as the practice spreads (→ Zoom Forms SWAT Team to Shape LLM Descriptions of Company).
• The Onion has reached a licensing agreement to relaunch Infowars as a satire site pending court approval — proceeds would flow to Sandy Hook judgment creditors .

Active questions and open splits.

• Section 230's design-versus-content boundary. The Massachusetts SJC has drawn the line; Meta argues it is legally meaningless. Federal courts handling the 30-state complaint wave will determine whether the distinction holds — and whether it extends beyond addiction claims to AI-generated content features .
• Direct AI-developer liability standard for content generation harms. The xAI and Google federal suits test whether inadequate safeguards against CSAM generation or manipulative chatbot behavior constitute actionable design defects — no settled standard exists, and discovery on internal safety protocols will be the battlefield .
• Scope and preemption of state chatbot disclosure statutes. Washington's HB 2225 is the most prescriptive enacted statute, but the West Coast cluster (California, Oregon, Washington) has different coverage scopes, exemptions, and enforcement mechanisms — and no federal floor exists to harmonize them (→ Washington Gov. Ferguson Signs HB 2225 Requiring AI Companion Chatbot Disclosures).
• Cross-border content moderation enforcement without U.S. cooperation. The DOJ's refusal to assist France on the X investigation means EU and national regulators will act unilaterally on algorithm manipulation and AI-generated illegal content claims — the enforcement gap is structural, not episodic .
• AI detector false positives as defamation exposure. If institutions act on AI-authorship determinations that are wrong, the defamation and wrongful termination exposure is real and unresolved — no liability standard for AI detection tool operators has been established .
• Corporate LLM-output management and disclosure obligations. Zoom's practice of submitting corrections to AI operators and optimizing public content to shape LLM descriptions raises questions about whether systematic corporate influence over AI outputs requires disclosure — and whether it could constitute deceptive trade practice if undisclosed (→ Zoom Forms SWAT Team to Shape LLM Descriptions of Company).
• Meta's plaintiff-recruitment ad ban and mass tort client sourcing. Whether Meta's removal of plaintiff-firm ads constitutes permissible terms-of-service enforcement or an actionable interference with attorney-client formation is unresolved — and whether other platforms adopt the same policy will determine the practical impact on mass tort litigation pipelines .

What to watch.

• Early motions practice in the xAI (Grok/CSAM) and Google (Gemini/suicide coaching) federal suits — whether courts entertain design-defect theories and what safety-protocol discovery they compel .
• Whether additional state AGs publish chatbot disclosure statutes following Washington's HB 2225 template, and whether any federal preemption challenge is filed before the January 1, 2027 effective date (→ Washington Gov. Ferguson Signs HB 2225 Requiring AI Companion Chatbot Disclosures).
• Whether France proceeds with unilateral enforcement against X executives following the DOJ's refusal to cooperate — and whether the EU Commission coordinates or acts independently .
• The Massachusetts AG's Meta addiction case at trial — whether the design-versus-content Section 230 distinction survives appellate scrutiny and how federal courts in the 30-state MDL respond .
• Whether the Stanford/Imperial/Internet Archive research team deploys its continuous monitoring tool, which would provide ongoing benchmarks for AI-content-share claims in platform governance and content authenticity litigation (→ Stanford study finds 35% of new websites AI-generated by May 2025).
• Court ruling on The Onion's Infowars licensing agreement — precedent for media platform disposition tied to defamation judgments and creditor recovery .