Erickson v. OpenAI: When Users Sue Over Their Own Uploads

• Erickson v. OpenAI presents a distinctive "user-upload" theory of AI copyright infringement, where the plaintiff alleges OpenAI infringed content he voluntarily uploaded into ChatGPT—a fundamentally different paradigm from the mass web-scraping cases dominating headlines.

• The case demonstrates the procedural power of forum-selection clauses, with OpenAI successfully transferring the case from Nevada to San Francisco federal court within weeks based on the ChatGPT Terms of Use.

• The complaint and key motion papers remain sealed, raising questions about how courts will balance public access rights against claims of proprietary or sensitive content in AI litigation.

• As the first prominent "terms-and-uploads" AI copyright case, Erickson may establish important precedents on how platform terms allocate rights in user inputs and outputs.

A Different Kind of AI Copyright Case

In the landscape of U.S. AI copyright litigation, most attention has focused on consolidated cases alleging that AI companies used millions of books, articles, and images from the internet to train their models. The Southern District of New York's multidistrict litigation against OpenAI and Microsoft exemplifies this paradigm. But Erickson v. OpenAI opens an entirely different front—one centered not on web-scale data acquisition but on a single user's allegation that OpenAI infringed content he himself uploaded into ChatGPT.

Michael Erickson, proceeding pro se, filed his lawsuit in Nevada state court in late 2025. The complaint was filed under temporary seal, leaving the precise claims shrouded. What we know comes primarily from the Nevada federal court's transfer order: Erickson alleges that OpenAI infringed content he uploaded into ChatGPT and that OpenAI breached its own Terms of Use in handling that content. This framing immediately distinguishes the case from other AI copyright disputes and raises contract-adjacent issues that major litigation has not squarely addressed.

Procedural Developments: Forum Selection in Action

OpenAI moved swiftly after removal to federal court, filing motions to seal, to transfer venue, and to stay proceedings pending transfer. The company's transfer motion relied on the forum-selection clause in the ChatGPT Terms of Use, which provides that all claims arising out of or relating to the Terms "will be brought exclusively in the federal or state courts of San Francisco, California." Judge Jennifer A. Dorsey of the District of Nevada granted transfer on February 10, 2026, treating Erickson's failure to oppose as consent under local rule.

The rapid enforcement of OpenAI's forum-selection clause tracks well-established Supreme Court doctrine. Under Atlantic Marine Construction Co. v. U.S. District Court (2013), forum-selection clauses pointing to another federal forum are ordinarily enforced through transfer, and courts give such clauses significant weight absent extraordinary circumstances. For AI companies with national user bases, this represents a significant litigation-management tool—disputes can be funneled to a single jurisdiction regardless of where users file.

The User-Upload Theory: Contractual Dimensions

The "user-upload" theory underlying Erickson's claims raises distinctive legal questions. When a user voluntarily submits content to an AI platform, what rights do they retain, and what uses has the platform licensed? OpenAI's current Terms of Use state that users retain ownership of their Input and own the Output (with ownership assigned by OpenAI), but OpenAI may use Content "to provide, maintain, develop, and improve" its services. The Terms also offer an opt-out mechanism for training use.

This language functions as a broad license for operational use and, unless opted out, for model improvement. An infringement claim based on user-uploaded content must therefore navigate this contractual framework: did the plaintiff opt out of training use, did the alleged infringement exceed the scope of licensed uses, or is the plaintiff challenging uses that fall outside any reasonable reading of the license grant? The closest analogous precedent may be A.V. v. iParadigms (2009), where the Fourth Circuit held that Turnitin's archiving of student papers for plagiarism detection constituted fair use—though the generative AI context raises additional complexities.

Sealing and Public Access Questions

The sealing of Erickson's complaint introduces another dimension that may become increasingly significant in AI copyright disputes. The Nevada court explicitly deferred long-term sealing decisions to the transferee court, maintaining seal on the complaint and transfer motion. Under Ninth Circuit law, sealing of judicial records requires specific justifications for departing from the presumption of public access. If Erickson's complaint contains sensitive chat logs or proprietary prompts, the Northern District of California will need to balance these interests against the public's right to understand what claims are being litigated.

Given that AI chat logs are increasingly central evidence in copyright disputes—as evidenced by the SDNY MDL's proceedings involving millions of ChatGPT logs—sealing battles may become a recurring feature of this litigation category. Whether courts permit AI copyright disputes to proceed largely under seal could limit the precedential transparency that helps shape industry practices and user expectations.

Our Mind

Erickson v. OpenAI may appear procedurally narrow, but it signals a litigation frontier that AI platforms and enterprise users should monitor closely. The case demonstrates that not all AI copyright disputes fit the "mass training data" mold; some will arise from the ordinary commercial relationship between platforms and their users, governed by the terms of use that users accept when they sign up.

For AI companies, the case validates the importance of well-drafted forum-selection and content-use provisions. OpenAI's ability to transfer the case within weeks of removal suggests that clear, enforceable terms can significantly shape litigation economics before any merits adjudication occurs. Companies should ensure that their terms accurately reflect their actual data handling, training practices, and opt-out mechanisms—any mismatch between policy language and operational reality becomes discovery risk.

For users—particularly businesses uploading proprietary or creative content to AI platforms—Erickson underscores the importance of understanding platform terms before uploading sensitive material. The content you upload may support later infringement claims, but it also triggers platform license terms that may authorize broad internal use and model improvement unless you affirmatively opt out. Users who value confidentiality should treat AI prompts like disclosure to a service provider and understand the scope of permissions they grant.

Finally, if courts hold that agreeing to platform terms licenses all complained-of uses, it would significantly narrow the scope of user infringement claims. If, on the other hand, courts find that certain uses exceed contractual permissions, it could open a new category of liability for AI platforms even when dealing with voluntarily submitted content. The initial case management conference is scheduled for May 14, 2026, making this a case to watch.

Key Takeaways

• Erickson v. OpenAI presents a "user-upload" theory of AI copyright infringement, distinct from mass web-scraping allegations in consolidated cases.

• OpenAI successfully enforced its forum-selection clause to transfer the case from Nevada to San Francisco within weeks, demonstrating the procedural power of platform terms.

• The complaint and key motion papers are sealed, raising questions about public access rights versus proprietary content protection in AI disputes.

• The case will likely turn on the scope of OpenAI's Terms of Use, including whether the plaintiff opted out of training use and whether alleged infringement exceeded licensed uses.

• A.V. v. iParadigms (the Turnitin case) provides the closest precedent for user-upload theories, though its fair use analysis may not map directly to generative AI.

• Evidence in AI copyright cases increasingly centers on chat logs, prompts, and outputs—creating preservation duties even in single-plaintiff cases.

• The Ninth Circuit's sealing standards will govern whether Erickson's complaint can remain under seal in the transferee court.

• For AI companies, the case validates clear forum-selection clauses, accurate content-use disclosures, and defensible opt-out mechanisms.

• For users, the case underscores the need to understand platform terms before uploading proprietary or creative content.

• The initial case management conference is scheduled for May 14, 2026, making this a case to watch for early rulings on user-upload AI copyright claims.

References

1. U.S. District Court for the District of Nevada, Case No. 2:26-cv-00083-JAD-BNW, Transfer and Sealing Order (Feb. 10, 2026)

2. U.S. District Court for the Northern District of California, Case No. 3:26-cv-01303-LJC, Scheduling Order (Feb. 12, 2026)

3. Justia Dockets & Filings, Erickson v. OpenAI Docket

4. OpenAI Terms of Use (current version)

5. Atlantic Marine Construction Co. v. U.S. District Court, 571 U.S. 49 (2013)

6. The Bremen v. Zapata Off-Shore Co., 407 U.S. 1 (1972)

7. A.V. ex rel. Vanderhye v. iParadigms, LLC, 562 F.3d 630 (4th Cir. 2009)

8. Kamakana v. City and County of Honolulu, 447 F.3d 1172 (9th Cir. 2006)

9. Center for Auto Safety v. Chrysler Group LLC, 809 F.3d 1092 (9th Cir. 2016)

10. In re OpenAI, Inc. Copyright Infringement Litigation (MDL 3143), S.D.N.Y.

SEO Keywords

Primary: Erickson v OpenAI, AI copyright litigation, user upload copyright, ChatGPT terms of use, forum selection clause AI

Secondary: AI platform terms, copyright infringement ChatGPT, AI user content rights, sealed complaint AI lawsuit, generative AI legal issues

Long-tail: OpenAI lawsuit 2026, AI training opt-out, platform content license, AI evidence discovery, Northern District California AI case