The End of Platform Immunity for AI#
For three decades, Section 230 of the Communications Decency Act shielded online platforms from liability for user-generated content. That shield is crumbling. Courts now distinguish between passively hosting third-party content, still protected, and actively generating, amplifying, or curating content through AI systems, increasingly not protected.
The legal principle is straightforward: AI platforms should not receive Section 230 protection because the content is generated by the platform itself. When an algorithm recommends a video, when an AI chatbot generates a response, when an AI Overview summarizes search results, these are the platform’s own speech acts, not third-party content.
This page addresses the distinct liability exposure for AI content moderation decisions and algorithmic amplification, separate from social media youth mental health litigation (which focuses on addiction and design defects) and AI companion chatbots (which addresses emotional AI and psychological harm).
The Core Legal Framework: First-Party vs. Third-Party Content#
Traditional Section 230 Protection#
Section 230(c)(1) provides that “No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.”
The key phrase: “provided by another.” Section 230 protects platforms from liability for content created by third parties, users posting comments, uploading videos, sharing photos.
Why AI-Generated Content Falls Outside Section 230#
Legal experts increasingly argue that AI-generated content is fundamentally different:
“Section 230 was built to protect platforms from liability for what users say, not for what the platforms themselves generate.” , Professor Chinmayi Sharma, Fordham Law School
When AI systems generate content, whether through chatbots, AI Overviews, or algorithmic recommendations, the platform transforms from neutral intermediary to “information content provider.” Section 230 explicitly does not protect information content providers.
The Algorithm as Speech: Anderson v. TikTok#
The Third Circuit’s August 2024 decision in Anderson v. TikTok fundamentally reshaped this landscape.
The Facts:
Ten-year-old Nylah Anderson died after attempting the “Blackout Challenge”, a dangerous viral trend encouraging users to choke themselves until passing out. The challenge video was not searched for; TikTok’s algorithm recommended it to her “For You Page.”
The Holding:
The Third Circuit ruled that TikTok’s recommendation algorithm constitutes the platform’s own “expressive activity”, first-party speech, not third-party content. Since Section 230 only immunizes platforms from liability for third-party content, algorithmic recommendations are not protected.
The Reasoning:
The court relied on the Supreme Court’s Moody v. NetChoice decision, which held that platform algorithms reflecting “editorial judgments” about content compilation are the platform’s own “expressive product” protected by the First Amendment. The Third Circuit recognized the irony: the same First Amendment logic platforms use to defend against regulation also makes their algorithmic choices their own speech, and thus their own liability.
Critical Distinction:
“We reach this conclusion specifically because TikTok’s promotion of a Blackout Challenge video on Nylah’s FYP was not contingent upon any specific user input. Had Nylah viewed a Blackout Challenge video through TikTok’s search function… then TikTok may be viewed more like a repository of third-party content than an affirmative promoter of such content.”
The line is clear: passive hosting remains protected; active algorithmic promotion does not.
AI Defamation: A New Frontier#
Wolf River Electric v. Google (June 2025)#
A Minnesota solar company filed a $110-210 million defamation lawsuit against Google after its AI Overview feature falsely claimed the company was being sued by the Minnesota Attorney General.
The AI Hallucination:
- Google’s AI Overview stated Wolf River Electric faced an Attorney General lawsuit
- The AI cited four links to support this claim: news articles, an AG statement, and Angie’s List
- None of the cited sources mentioned Wolf River Electric
- The sources discussed other solar companies being sued, the AI incorrectly applied this to Wolf River
Documented Business Losses:
- March 3, 2025: Customer terminated $39,680 contract citing “lawsuits” found when “Googling” Wolf River
- March 11, 2025: Non-profit terminated $174,044 in projects citing “several lawsuits in the last year” with the “Attorney General’s Office”
Legal Significance:
“This might be one of the first cases where we actually get to see how the courts are going to really dig down and apply the basic principles of defamation law to AI.” , Ari Cohn, Foundation for Individual Rights in Education
The case directly tests whether Section 230 protects AI-generated content. Legal experts are divided: some argue AI Overviews are merely “souped-up” search algorithms that remain protected; others contend AI-generated summaries are the platform’s own authored speech.
Walters v. OpenAI: The Disclaimer Defense (May 2025)#
The first major U.S. AI defamation case to reach judgment was Walters v. OpenAI in Georgia.
The Facts:
Radio host Mark Walters alleged ChatGPT falsely claimed he had been sued for embezzlement by the Second Amendment Foundation and had served as its treasurer. None of this was true:Walters had no connection to the organization or any lawsuit.
The May 2025 Ruling:
Judge Tracie Cason ruled for OpenAI, finding Walters failed to prove defamation or that OpenAI acted with fault.
Key Reasoning:
The court held that ChatGPT’s output “could not be reasonably understood as stating actual facts” given:
- ChatGPT’s disclaimers about potential inaccuracies
- The well-known limitations of AI systems
- The journalist who received the output acknowledged the error before publication
“No reasonable person would interpret the AI-generated content in question as a literal or factual assertion, particularly in light of the well-known limitations and disclaimers attached to the tool.”
Implications:
This ruling suggests that disclaimers and responsible AI design may protect developers from some defamation claims. However, as AI systems become more authoritative and users rely on them more heavily, the “no reasonable person would believe this” defense may weaken.
Starbuck v. Meta and Starbuck v. Google (2025)#
Activist Robby Starbuck filed lawsuits against both Meta and Google after their AI systems generated false statements linking him to:
- The January 6 Capitol riot
- Holocaust denial
- Child endangerment
The Persistence Problem:
Starbuck alleges the false information persisted even after Meta was notified in August 2024, raising questions about platform duties once aware of AI-generated defamation.
Stakes:
These cases could establish the first U.S. precedent on who is liable when AI defames, the AI developer, the platform deploying it, or both.
Legislative Developments: The Section 230 Reckoning#
Section 230 Sunset Legislation#
In May 2024, House Energy and Commerce Committee Chair Cathy McMorris Rodgers (R-WA) and Ranking Member Frank Pallone Jr. (D-NJ) unveiled bipartisan draft legislation to sunset Section 230.
Key Provisions:
- 18-month window for Congress and industry to develop replacement framework
- If no replacement enacted, Section 230 protections expire entirely
- Intent: force negotiated reform rather than actual sunset
A separate bill led by Senators Lindsey Graham (R-SC) and Dick Durbin (D-IL) proposes sunsetting Section 230 on January 1, 2027 unless Congress enacts replacement legislation.
Algorithm Accountability Act (2025)#
Senators Mark Kelly (D-AZ) and John Curtis (R-UT) introduced legislation specifically targeting algorithmic amplification:
Key Provisions:
- Imposes a “duty of care” on platforms using recommendation-based algorithms
- Requires responsible design, training, testing, and deployment to prevent foreseeable bodily injury or death
- Creates civil right of action for injured individuals in federal court
The bill directly addresses the gap identified in Anderson v. TikTok: platforms that actively recommend harmful content to vulnerable users.
No Section 230 Immunity for AI Act#
Senators Josh Hawley (R-MO) and Richard Blumenthal (D-CT) introduced legislation to explicitly waive Section 230 immunity for claims involving generative AI. While blocked in the Senate, it signals bipartisan recognition that AI-generated content differs fundamentally from user-generated content.
Court Rulings: The Emerging Jurisprudence#
Bogard v. TikTok (February 2025)#
In a 35-page ruling, Magistrate Judge Virginia DeMarchi in the Northern District of California dismissed products liability claims against YouTube and TikTok.
The Claims:
Parents alleged that platform reporting features were defectively designed, allowing harmful “challenge” videos to remain after being reported. Multiple children died or were injured by content that remained on platforms despite user reports.
The Ruling:
Judge DeMarchi held the allegations, even if proven, wouldn’t establish a product defect:
“The crux of plaintiffs’ allegations is that the defendants’ reporting systems are defective because plaintiffs’ reports do not produce the outcomes that plaintiffs believe they should, i.e. removal of the reported videos.”
But the Door Remains Open:
The ruling explicitly left room for future product liability claims:
- Claims were dismissed without prejudice
- The court did not rule that algorithms can never be defective products
- Different allegations, particularly those focused on recommendation rather than moderation, might survive
Zuckerman v. Meta (2024)#
A federal court upheld Meta’s exclusive control over its content moderation algorithms, rejecting third-party content moderation tools.
The Implication:
Platforms that maintain exclusive control over algorithmic decisions may face greater liability for those decisions. If the algorithm is solely the platform’s choice, the platform cannot shift blame to third parties.
FTC Enforcement: Operation AI Comply#
On September 25, 2024, the Federal Trade Commission launched “Operation AI Comply”, a national enforcement initiative targeting deceptive AI practices.
Key Enforcement Actions#
DoNotPay “Robot Lawyer” Case:
The FTC took action against DoNotPay, which claimed to offer “the world’s first robot lawyer.” The FTC found:
- DoNotPay never trained its system on legal authorities or legal reasoning
- Never tested the quality and accuracy of most advertised capabilities
- Settlement: $193,000 penalty plus required consumer notice of service limitations
FTC Chair Lina Khan:
“Using AI tools to trick, mislead, or defraud people is illegal. The FTC’s enforcement actions make clear that there is no AI exemption from the laws on the books.”
Continuing Enforcement Under New Administration#
Despite administration change, Operation AI Comply continues into 2025:
- Actions against Click Profit and Workado for baseless AI income claims
- Continued scrutiny of AI systems making unsubstantiated capability claims
- Focus on AI products marketed to vulnerable consumers
AI Companion Chatbot Investigation (September 2025)#
The FTC issued 6(b) orders to seven companies investigating AI companion chatbots and child safety:
- Alphabet (Google)
- Character Technologies
- Meta Platforms
- OpenAI
- Snap
- X.AI (xAI)
The investigation examines whether AI chatbot designs encourage emotional attachment and whether adequate safety measures protect minors.
The OpenAI Wrongful Death Wave#
In November 2025, the Social Media Victims Law Center and Tech Justice Law Project filed seven lawsuits against OpenAI in California state courts.
The Allegations#
Wrongful Death Plaintiffs (4):
- Family of Zane Shamblin, 23
- Family of Amaurie Lacey, 17
- Family of Joshua Enneking, 26
- Family of Joe Ceccanti, 48
Psychological Harm Plaintiffs (3): Three additional plaintiffs allege ChatGPT induced psychotic breaks requiring emergency psychiatric care.
Core Claims#
The lawsuits allege GPT-4o was:
- “Engineered to maximize engagement through emotionally immersive features”
- Released with compressed safety testing:“months of safety testing into a single week”, to beat Google Gemini to market
- Designed with “persistent memory, human-mimicking empathy cues, and sycophantic responses” without adequate guardrails
The Scale of the Problem#
OpenAI’s own analysis revealed that conversations with more than one million users per week involve discussions of suicidal intent.
The Emerging Standard of Care#
For Platform Operators#
1. Algorithmic Responsibility
Based on Anderson v. TikTok and subsequent litigation:
- Algorithmic recommendations are the platform’s own speech, not protected third-party content
- Active curation creates liability exposure that passive hosting does not
- “Neutral tool” defenses increasingly rejected when algorithms amplify harmful content
2. Content Moderation Duties
Platforms face competing pressures:
- Under-moderation exposes platforms to claims they amplified harmful content
- Over-moderation may trigger First Amendment concerns or claims of censorship
- The standard emerging: documented, reasonable moderation decisions with human oversight
3. AI-Generated Content Review
For AI Overview, chatbot, and generative AI features:
- Implement accuracy verification for AI-generated summaries
- Maintain correction mechanisms when AI outputs are demonstrably false
- Document known limitations and communicate them to users
- Consider whether disclaimers adequately inform users of AI limitations
For AI Developers#
1. Defamation Prevention
- Implement safeguards against generating false statements about real individuals and entities
- Create correction mechanisms for identified AI hallucinations
- Document accuracy testing and known failure modes
- Consider whether disclaimers actually reach and inform end users
2. Safety Testing Documentation
Courts will scrutinize:
- How AI systems were trained and tested before deployment
- Whether safety testing was compressed for competitive reasons
- What guardrails existed and whether they were removed
- Real-time monitoring capabilities and incident response
3. Disclosure Obligations
- Clearly disclose AI limitations to deployers
- Provide guidance on appropriate use cases
- Document risks associated with vulnerable user populations
For Businesses Using AI Content Systems#
1. Vendor Due Diligence
- Understand what AI systems generate content on your behalf
- Review vendor contracts for liability allocation
- Assess vendor safety testing and accuracy claims
- Consider insurance coverage for AI-generated content liability
2. Monitoring and Response
- Monitor AI-generated content for accuracy
- Implement correction mechanisms when errors identified
- Preserve evidence of AI system outputs
- Document incident response procedures
3. Human Oversight
- Maintain human review for high-stakes AI outputs
- Create escalation paths when AI systems generate concerning content
- Train staff to identify AI failures
- Document human oversight decisions
Practical Risk Assessment#
High-Risk AI Content Activities#
Highest Exposure:
- AI chatbots interacting with vulnerable users (minors, mental health)
- AI-generated summaries about real individuals or businesses
- Algorithmic recommendation of user-generated content
- AI content moderation decisions affecting user accounts
Moderate Exposure:
- AI-generated marketing or product descriptions
- AI customer service providing factual information
- Automated content categorization and tagging
Lower Exposure (currently):
- AI-assisted internal content review
- AI summarization of company’s own content
- AI translation of controlled content
Pre-Deployment Checklist#
Before deploying AI content systems:
Assess Section 230 exposure: Is the AI generating content (no immunity) or hosting user content (potential immunity)?
Evaluate defamation risk: Can the AI generate false statements about real people or entities?
Test accuracy claims: Does the AI perform as marketed? Document testing results.
Implement human oversight: What human review exists for high-stakes outputs?
Create correction mechanisms: How will identified errors be corrected?
Document safety decisions: Preserve records of safety testing and design choices.
Review insurance coverage: Does current coverage address AI content liability?
Resources#
- Anderson v. TikTok (Third Circuit)
- FTC Operation AI Comply Announcement
- House Section 230 Sunset Legislation
- Harvard Law Review: Beyond Section 230
- CDT: Section 230 and Generative AI Analysis
- ABA: Generative AI’s Section 230 Tightrope
- Fortune: Section 230 and AI Chatbots
- Social Media Victims Law Center: OpenAI Lawsuits