Northern District Applies Section 230 to AI-Assisted Content Moderation

In Ryan v. X Corp., a Northern District of California court held that Section 230 of the Communications Decency Act immunized X (formerly Twitter) against claims arising from suspension of a user’s account, notwithstanding that X allegedly used an AI system to select the account for suspension. The case also illustrates the unsettled state of the law (at least in the Ninth Circuit) regarding whether contract-based claims based on a platform’s alleged failure to comply with its own terms of service regarding content moderation are excluded from the Section 230 safe harbor.

Plaintiff Jeremy Ryan, a “successful NFT artist” who used X to “further develop his cryptocurrency and NFT community,” sued X for suspending his accounts without explanation. Ryan alleged that X used an AI system to select his account for suspension. Ryan’s claims included breach of contract, promissory estoppel, unjust enrichment, intentional interference with contractual relationships with third parties, interference with prospective economic advantage, promissory fraud, conversion, and unfair business practices.

The opinion in this case came in two parts. The court first dismissed several of Ryan’s claims in September 2024, holding that the breach of contract, intentional interference, conversion, and unfair business practices claims were barred by Section 230. The court then dismissed the remaining claims in a subsequent opinion in December 2024 as barred by X’s terms of service.

In its September opinion, the court applied the three factors from Barnes v. Yahoo! to determine whether the statute immunized X from liability: (1) whether X is a provider or user of an “interactive computer service,” (2) whether Ryan’s claim seeks to treat the defendant as a “publisher or speaker,” and (3) whether the content at issue was provided by “another information content provider.” The court quickly determined that the first and third factors were satisfied: X is an interactive computer service, and Ryan provided the content at issue. The key question was whether X’s suspension of Ryan’s accounts constituted a “publishing decision.”

The court ruled that X’s account suspension decisions were indeed “traditional publishing functions” protected under Section 230. Relying on Ninth Circuit precedent, the court held that “any activity that can be boiled down to deciding whether to exclude material that third parties seek to post online is perforce immune under section 230.” The fact that X may have used AI to assist in its moderation decisions did not change this outcome.

Other than the AI angle, this would appear to be a fairly standard application of Section 230, but the court also had to address Ryan’s breach of contract claim based on X’s alleged failure to comply with its own terms of service and other statements made on X’s website. The court’s job was made more difficult by recent Ninth Circuit cases Calise v. Meta Platforms, Inc. and Estate of Carson Bride v. YOLO Technologies, Inc., both of which found exceptions to Section 230 based on the defendant platforms’ alleged failure to comply with their own content moderation policies.

Commentators have criticized Calise and YOLO for significantly expanding the limited Section 230 exception for promissory estoppel claims in Barnes v. Yahoo! in a manner that is inconsistent with dozens of prior cases and with the words of the statute itself. Barnes created a limited exception to Section 230 for contract-based claims where the defendant platform operator made a specific promise to the plaintiff to remove content and then failed to do so. Calise and YOLO expanded the exception to include “promises” contained in generally applicable platform terms and policies.

The Ryan court held that these cases did not preclude application of Section 230 and reached back to Barnes to support its conclusion that Ryan’s contract claim was barred, holding that “despite Ryan’s protestations to the contrary, his suspension was undertaken by X in its capacity as a publisher of information, it meets the other Barnes criteria, and is protected.” In discussing Calise, the court noted that Ryan was trying to “impose liability upon X arising from its actions as a publisher, not its failure to fulfill some promise separate from that role,” and that the argument that his contract claim arose from a separate promise that X made to users on its website “comes up short because . . . statements made in a social media site’s terms and community standards do not amount to a legally enforceable promise.” We leave it to the reader to decide whether Ryan actually distinguishes Calise or simply refuses to apply it.

The court went on to dispose of Ryan’s claims for intentional interference with plaintiff’s contractual relationships, conversion, and unfair business practices, also on Section 230 grounds. In each instance, Ryan’s claim arose from X’s decision to suspend Ryan’s accounts, which the court held was a publishing decision and therefore immunized under Section 230. In discussing Ryan’s conversion claim, which was based on the allegation that an “automated bot” made certain misrepresentations to Ryan regarding restoration of his account, the court noted that suspending an account is a publishing decision, regardless of whether AI is used.

Ryan did not fare any better in the December opinion, where the court dismissed his remaining claims as barred by X’s terms of service. X’s terms of service stated that X may stop providing services to users, limit use of services, remove user content, and suspend or terminate users without liability. Ryan argued that California courts do not enforce limitation of liability provisions under California Civil Code Section 1668 if they protect against fraud or gross negligence. Only Ryan’s claim for promissory fraud alleged fraud or gross negligence, however, and that claim requires a promise. Because X did not make any promise not to suspend Ryan’s accounts even if he complied with the terms of service, California Civil Code Section 1668 did not prohibit enforcement of the limitation of liability in X’s terms of service. Accordingly, the court dismissed Ryan’s claims for promissory fraud, promissory estoppel, unjust enrichment, and unfair business practices claim.

The court also noted that, even if not barred by X’s terms of service, Ryan’s unjust enrichment claim would be barred under Section 230. Ryan argued that X was not entitled to Section 230 immunity because the decision to suspend his account was made by X’s AI system, and that this somehow made X an information content provider rather than merely an interactive computer service provider. Ryan also argued that X’s AI system (as distinct from X) was not entitled to Section 230 immunity. Ryan’s arguments are difficult to understand based on the opinion, and it appears that the court struggled with them as well. In any event, the court was not convinced, stating “no authority suggests that X’s use of a generative AI model to do content regulation deprives it of Section 230 immunity” and “[n]othing in the plain language of the statute suggests that X would be denied section 230 protection because it uses an AI system for content moderation.”