Section 230 & Generative AI
The Statutory Framework
47 U.S.C. § 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 question: Are LLM outputs provided by another information content provider (the user), or is the AI company itself the information content provider?
The Material Contribution Test
Fair Housing Council v. Roommates.com (9th Cir. 2008)
Courts apply the "material contribution" test to determine if an online service is an information content provider. A service that "materially contributes" to the allegedly illegal content cannot claim Section 230 immunity.
"If we hold that only those defendants who are responsible for every single element of the creation or development of illegal information are responsible in part for the creation and development of that information, the phrase 'in part' becomes meaningless."2
Application to LLMs
Most legal scholars and the co-authors of Section 230 itself agree that generative AI outputs likely fall outside Section 230 protection.
Senator Ron Wyden (co-author): "Section 230 does not protect generative AI outputs."3
Matt Perault (Duke): "It is difficult to imagine that a court would find [LLMs do not develop content] if an LLM drafts text that would otherwise be illegal."4
Key Arguments
Against Section 230 Immunity
- Novel Content Creation: LLMs generate new text, not merely transmit user content
- Material Contribution: AI companies train models, design prompts, and shape outputs
- Not "Another" Provider: When an AI creates defamatory content, it's the AI company's creation
- Hallucinations: AI-generated falsehoods are not "provided by another information content provider"
For Section 230 Immunity (Minority View)
- User Prompts: AI merely responds to user queries, like a sophisticated search engine
- Interactive Computer Service: AI systems clearly qualify as interactive computer services
- Neutral Tool: AI is a tool that users direct; outputs are user-directed content
Litigation Signals
Walters v. OpenAI & Battle v. Microsoft
Notably, neither OpenAI nor Microsoft raised Section 230 as a defense in recent defamation cases5 involving ChatGPT and Bing AI outputs. This suggests AI companies themselves may not believe Section 230 provides viable protection.
"The companies' decision not to assert Section 230 in these cases may signal how they anticipate the analysis would have played out."6
The "Verbatim" Exception
Courts may recognize an exception when AI outputs are so close to user prompts that they constitute "verbatim" transmission rather than creation. In such cases, Section 230 might apply.
Example: If a user prompts "Write exactly: [defamatory statement]" and the AI reproduces it verbatim, the AI company might argue it did not "materially contribute" to the content.
However, this exception would not apply to hallucinations or novel AI-generated content.
Summary Table
| AI Use Case | Section 230 Likely? | Reasoning |
|---|---|---|
| Novel AI-generated content | NO | AI company is information content provider |
| AI hallucinations/false statements | NO | Not "provided by another" |
| Verbatim user prompt reproduction | MAYBE | May qualify as neutral transmission |
| Search/retrieval of existing content | YES | Traditional 230 protection applies |
Sources
Analysis drawn from: CDT Section 230 Analysis (2024), Matt Perault "Section 230 Won't Protect ChatGPT" (J. Free Speech L. 2023), Harvard Law Review "Beyond Section 230" (2023), and Congressional Research Service LSB11097.