The Central Question#
Does Section 230 of the Communications Decency Act:“the 26 words that created the internet”, protect AI companies from liability for content their systems generate?
This question has massive implications for AI deployers. If Section 230 provides immunity, companies can deploy chatbots, AI assistants, and generative systems with reduced legal exposure. If it doesn’t, every AI-generated output becomes a potential source of liability.
The emerging consensus: Section 230 likely does not protect AI-generated content. But courts haven’t definitively ruled, and the legal landscape is shifting rapidly.
What Section 230 Says#
Section 230(c)(1) provides:
“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.”
This language immunizes platforms from liability for content created by third parties (users). The immunity has been interpreted broadly by courts, protecting social media companies from defamation claims, content moderation decisions, and algorithmic amplification of user content.
The key phrase is “provided by another information content provider.” Section 230 protection applies only to content created by someone else, not content the platform itself creates.
Why Section 230 May Not Apply to AI#
The Generation Problem#
“Section 230 was built to protect platforms from liability for what users say, not for what the platforms themselves generate,” explains Professor Chinmayi Sharma of Fordham Law School.
Traditional platforms host user content. AI chatbots generate new content. This is a fundamental distinction.
As Sharma observes: “Courts are comfortable treating [extraction and curation] as hosting or curating third-party content. But transformer-based chatbots don’t just extract. They generate new, organic outputs personalized to a user’s prompt. That looks far less like neutral intermediation and far more like authored speech.”
The “Information Content Provider” Analysis#
Under Section 230, an “information content provider” is “any person or entity that is responsible, in whole or in part, for the creation or development of information.”
Courts apply a “material contribution test”: Did the platform contribute significantly to the creation of the content? If so, the platform may be considered an “information content provider” itself, and lose Section 230 protection.
With generative AI, the platform’s code determines what gets communicated. As data-privacy lawyer Collin Walke argues: “From a pure textual standpoint, AI platforms should not receive Section 230 protection because the content is generated by the platform itself. Yes, code actually determines what information gets communicated back to the user, but it’s still the platform’s code and product, not a third party’s.”
The User Prompt Argument#
AI companies argue that user prompts make the user the “information content provider,” with the AI merely processing and responding to user-provided content.
This argument has weaknesses:
- The AI adds substantial content: A user asking “What is the capital of France?” doesn’t create the response “Paris”, the AI does
- Hallucinations are platform-generated: When ChatGPT invents false information, no user provided that content
- The platform controls the training and outputs: AI companies choose training data, model architecture, and safety systems
Key Court Decisions#
Anderson v. TikTok (Third Circuit, 2024)#
The Third Circuit’s decision in Anderson v. TikTok established that algorithmic recommendation can constitute “expressive activity” not protected by Section 230.
The Facts:
- A 10-year-old died after attempting TikTok’s “Blackout Challenge”
- The algorithm had recommended the dangerous content to her
- TikTok argued Section 230 barred the lawsuit
The Holding: The court ruled TikTok could face liability because it wasn’t merely hosting third-party content, it was “actively recommending specific content to users,” engaging in its own form of expression.
Significance for AI: If algorithmic curation can lose Section 230 protection, AI generation, which goes far beyond curation, is even more vulnerable.
Bogard v. TikTok / YouTube (N.D. Cal., February 2025)#
In February 2025, Magistrate Judge Virginia DeMarchi dismissed products liability claims against YouTube and TikTok related to “choking challenge” videos.
The Facts:
- Parents sued after children died attempting challenges seen on the platforms
- They alleged the platforms’ reporting systems were defectively designed
- Claims included products liability, negligence, and misrepresentation
The Ruling: Judge DeMarchi dismissed the claims, finding plaintiffs failed to clearly identify the “product” or “design defect.” She also held Section 230 barred the products liability claims.
Key Limitation: This case involved platforms’ failure to remove user-generated content, the traditional Section 230 scenario. It doesn’t address liability for content the platform itself generates.
Plaintiffs given leave to amend, suggesting the door isn’t fully closed.
Walters v. OpenAI (Georgia State Court, 2023-2024)#
Radio host Mark Walters sued OpenAI after ChatGPT falsely claimed he had been “accused of defrauding and embezzling funds” from a nonprofit.
OpenAI’s Motion Denied: The Georgia state court denied OpenAI’s motion to dismiss, allowing the defamation case to proceed.
Significance: While not a definitive ruling on Section 230, the court’s willingness to let the case proceed suggests AI companies can’t easily escape defamation liability.
Garcia v. Character.AI (Florida, 2024-2025)#
After a 14-year-old died by suicide following extensive chatbot interactions, his mother sued Character.AI.
Notable Absence: Character.AI has not invoked Section 230 as a defense.
As Pete Furlong of the Center for Humane Technology observed: “I think that that’s really important because it’s kind of a recognition by some of these companies that that’s probably not a valid defense in the case of AI chatbots.”
The Algorithmic Amplification Cases#
A related line of cases addresses whether platforms can be liable for algorithmic promotion of harmful content, a question relevant to AI systems that curate and recommend.
The “Neutral Tool” Framework#
Courts have historically applied a “neutral tools” analysis. In Force v. Facebook (2d Cir. 2019), the court held that Facebook’s content recommendation algorithms were “neutral tools” entitled to Section 230 protection because they treated all content similarly.
The Erosion of “Neutral Tools”#
Recent decisions have narrowed this protection:
Lemmon v. Snap (9th Cir. 2021): Parents sued Snapchat after teens died in a car crash allegedly caused by the app’s “speed filter” encouraging dangerous driving. The Ninth Circuit allowed the case to proceed, holding plaintiffs weren’t suing Snap as a “publisher” but for “violat[ing] its distinct duty to design a reasonably safe product.”
Anderson v. TikTok (3d Cir. 2024): As noted above, the court held algorithmic promotion could constitute “expressive activity” outside Section 230 protection.
The Social Media Addiction MDL#
The Adolescent Social Media Addiction litigation tests whether platforms can be liable for algorithmic designs allegedly harming children.
Scale of Litigation#
- 1,867 cases are currently pending in the MDL before Judge Yvonne Gonzalez Rogers (N.D. Cal.)
- Bellwether trials are scheduled for 2026
- State court proceedings are also advancing, with California trials beginning November 2025
Core Allegations#
Plaintiffs allege social media algorithms are designed to maximize engagement at the cost of children’s mental health, driving addiction, anxiety, depression, and suicide.
Relevance to AI#
These cases establish that:
- Algorithmic design choices can create liability beyond traditional content hosting
- Courts are willing to look behind Section 230 when platforms actively shape user experiences
- “The algorithm did it” is not necessarily a defense
The principles developed in social media addiction litigation will likely inform AI liability cases.
Legislative Developments#
No Section 230 Immunity for AI Act#
Senator Josh Hawley introduced legislation in 2023 to explicitly exclude generative AI from Section 230 protection. While the bill hasn’t passed, it signals Congressional interest in clarifying the law.
Algorithm Accountability Act (2025)#
Introduced November 19, 2025, this bill would create accountability for algorithmic harms, a “targeted fix” rather than wholesale Section 230 repeal.
State-Level Action#
States are actively legislating AI and algorithmic accountability:
- California has multiple bills addressing AI content and platform liability
- Utah’s AI Law explicitly provides AI use is not a defense to consumer protection violations
- Colorado’s AI Act creates obligations for deployers of high-risk AI systems
Practical Implications for AI Deployers#
What This Means#
If you deploy AI chatbots or generative systems:
Don’t assume Section 230 protection exists. Courts haven’t definitively ruled, but the weight of analysis suggests AI-generated content won’t be protected.
Treat AI outputs as potential liability sources. Every generated statement could form the basis for defamation, negligence, or other claims.
Implement content safety measures. Systems that prevent harmful outputs reduce liability exposure regardless of Section 230 status.
Document your safety efforts. Evidence of reasonable care may be relevant to negligence analysis.
The Defamation Risk#
AI hallucinations create particular defamation exposure:
- ChatGPT has falsely accused individuals of crimes
- AI systems have fabricated professional misconduct
- Generated content has attributed fake quotes to real people
If Section 230 doesn’t apply, traditional defamation elements, publication, falsity, fault, damages, apply to AI outputs.
The Harmful Content Risk#
Beyond defamation, AI systems face liability for:
- Suicide and self-harm content (Garcia, Raine cases)
- Dangerous instructions (how-to content for harmful activities)
- Privacy violations (disclosure of personal information)
- Discrimination (biased outputs in employment, housing, credit)
The Emerging Legal Framework#
Courts Will Lead#
Given the Congressional Research Service analysis: “Given these circumstances, it seems likely that courts, rather than legislators, will take the lead in defining the limits of Section 230’s applicability to generative AI technology.”
A Patchwork Is Likely#
Different circuits may reach different conclusions, creating uncertainty until the Supreme Court addresses the issue or Congress acts.
The Third Circuit’s Anderson decision represents one approach, treating algorithmic activity as potentially unprotected “expressive activity.” Other courts may adopt different frameworks.
The Key Distinction#
Courts are likely to distinguish between:
- Hosting/curating third-party content → Section 230 likely applies
- Generating new content → Section 230 likely doesn’t apply
AI systems that primarily extract and display user content may retain protection. Systems that synthesize, generate, and create new outputs likely won’t.
What AI Companies Are Doing#
Character.AI’s Approach#
By not invoking Section 230 in the Garcia case, Character.AI implicitly acknowledges the defense may not apply to chatbot outputs.
OpenAI’s Litigation Strategy#
OpenAI has faced multiple lawsuits without successfully invoking Section 230 as a complete defense. The company focuses instead on challenging specific elements of claims.
Industry Uncertainty#
The AI industry lacks a clear legal framework. Companies are:
- Strengthening content safety systems
- Improving monitoring and incident response
- Negotiating indemnification provisions with deployers
- Watching litigation developments closely
Conclusion#
The 26 words of Section 230 were written for a different era, when platforms hosted user bulletin boards, not AI systems that generate novel content. Courts and commentators increasingly recognize this distinction.
For AI deployers, the prudent approach is to assume Section 230 won’t provide protection and to implement safety measures, documentation, and oversight accordingly. The legal framework is evolving, and companies that treat AI outputs as potential liability sources, rather than hoping for statutory immunity, will be better positioned regardless of how courts ultimately rule.
Resources#
- CDT: Section 230 and its Applicability to Generative AI
- Stanford HAI: Does Section 230 Cover Generative AI?
- Congressional Research Service: Section 230 Immunity and Generative AI
- Harvard Law Review: Beyond Section 230 - Principles for AI Governance
- Lawfare: Section 230 Won’t Protect ChatGPT
- ABA: Beyond the Search Bar - Generative AI’s Section 230 Tightrope Walk
- Social Media Addiction MDL Updates