Spyder Games v. Mementum Lab: The First Federal Test of AI Character Copyright Is Underway

On November 26, 2025, two Roblox developers filed a lawsuit in federal court that will probably decide who owns AI-generated meme characters in the United States.

The case is Spyder Games LLC et al. v. Mementum Lab, 5:25-cv-10248, Northern District of California. The plaintiffs are SpyderSammy's Spyder Games LLC, the developer behind Steal a Brainrot (one of the most popular games in Roblox history with peak concurrent users above 25 million), and Speedy Simulator Gaming. The defendant is Mementum Lab, a Paris-based agency representing several Italian Brainrot creators including Noxa, the Indonesian TikToker who originated Tung Tung Tung Sahur.

The plaintiffs aren't suing for damages. They're asking the court for a declaratory judgment that AI-generated brainrot characters cannot be copyrighted at all. Mementum will argue the opposite. Whoever wins will set the precedent that determines whether thousands of viral AI characters become licensable corporate IP or remain free for anyone to use.

I built CharacterClaim.ai to assess these exact questions. This is the case that will reshape the answer. Here's what's actually being argued, in plain language.

The parties and the stakes

Spyder Games LLCis owned by Sammy (“SpyderSammy”), the developer of Steal a Brainroton Roblox. The game launched in May 2025 and became one of the platform's largest titles within months. By August 2025, it crossed 20 million concurrent users, briefly surpassing Fortnite's all-time concurrent record. Italian Brainrot characters are the core gameplay loop. Players collect, steal, and trade voxel versions of Tung Tung Tung Sahur, Ballerina Cappuccina, Tralalero Tralala, and many others.

Mementum Lab is a France-based agency and creator collective. According to public sources including their own representations, they license characters from the original creators (Noxa, Breno, alexey_pigeon, Tim Cheese) and represent the creators in commercial deals. Mementum positions itself as a legitimate IP holder for these characters, claiming ownership through licensing agreements with the human creators who prompted them into existence.

The commercial stakes are enormous. Epic Games launched Tung Tung Tung Sahur and Ballerina Cappuccina skins in Fortnite on April 1, 2025, licensing both characters from Mementum. Epic's senior communications manager Brian Sharon confirmed to GamesBeat that the skins are “licensed from Brainrot Factory.” Individual Fortnite skin releases have generated as much as $50 million in past years. The Mementum-licensed Steal the Brainrot island in Fortnite Creative hit 500,000 concurrent users at peak.

If Mementum's licensing theory holds, that revenue stream is legitimate and Mementum is positioned as the licensing gateway to the entire Italian Brainrot universe. If it doesn't, the characters are public domain and the Epic licensing deal was technically unnecessary.

The Spyder Games argument

The plaintiffs' position is straightforward and rests on settled U.S. copyright law.

Their argument, as summarized in the public docket and in attorney Aaron Moss's interview with GamesBeat: AI-generated images are not copyrightable in the United States without sufficient human authorship. The U.S. Copyright Office has held this position consistently since its March 2023 guidance, and the position was reinforced in the USCO's Copyright and Artificial Intelligence, Part 2: Copyrightability report (January 2025). The leading case, Thaler v. Perlmutter, was decided in favor of the Copyright Office at the D.C. Circuit in March 2025; the Supreme Court denied certiorari in March 2026. The human-authorship requirement is now judicially settled.

Moss told GamesBeat: “The Italian Brainrot characters that we're talking about here have all come out of AI image generators, and the copyright office has been very clear and very consistent in its guidance, which is that if something is generated by AI without sufficient human authorship, it's not copyrightable, and no court has disagreed with that.”

The plaintiffs' theory: a prompt is not enough. Per the USCO Part 2 report, a user who enters a text prompt into an AI image generator is not the author of the resulting image because the user does not control the specific expressive choices the AI makes. Selection of one output among several is also generally insufficient. Without substantial human modification of the AI output, no copyright attaches.

If that legal framework controls, Mementum has nothing to license. The characters are not copyrightable, so there's no IP to assign, and Mementum's licensing claims fail.

The Spyder Games argument also has a practical, almost public-interest framing. Moss again: “If we win, brainrot characters stay free for everyone, creators, game developers, memers and really the whole internet. If Mementum wins, it's like Marvel or Disney characters, where one company controls who can use all of those brainrots.”

The Mementum argument

The defendants' position has not been fully laid out in court filings yet, but their public theory is more nuanced than it might appear.

Mementum's argument, as reported by GamesBeat and consistent with their public representations: AI-generated meme characters can be copyrighted when they reflect substantial human creative input. Noxa didn't simply type a generic prompt and accept the first output. He coined a name with cultural resonance (the kentongan drum sound during Ramadan, the sahur meal, the mythology of a wooden figure who visits non-responsive sleepers). He developed the character's backstory, voice, and lore. He chose the specific outputs. The “character” as it exists in commerce is the cumulative product of human creative choices, not a single AI image.

Mementum points to historical precedent for meme creators owning IP in viral characters. Charles Schmidt successfully claimed and licensed Keyboard Cat (originally a 1984 video edited by Schmidt, popularized as a meme starting 2009). The Davies-Carr family monetized Charlie Bit My Finger and sold it as an NFT for over $700,000 in 2021. These weren't AI-generated, but Mementum's argument treats the underlying principle as the same: the creator of a viral character has a legitimate ownership claim, regardless of the specific medium of creation.

Mementum's stronger argument is the one anchored in the USCO's own guidance. The Copyright Office Part 2 report does not say AI works are categorically uncopyrightable. It says they're uncopyrightable to the extent the expressive elements were generated by AI without sufficient human authorship. The same report acknowledges that human-authored elements layered onto AI output can be independently protected, and that “selection, arrangement, and modification” of AI output by a human can produce protectable authorship. Per 17 U.S.C. §103(b), original elements added to an unprotectable work are protected separately from the underlying work.

Mementum will likely argue that the character of Tung Tung Tung Sahur, taken as a whole, includes substantial human authorship: the name, the lore, the cultural context, the manual selection and curation of outputs across hundreds of TikTok posts, the catchphrases, the mythology Noxa wrote. Even if any single AI-generated frame is uncopyrightable, the character as a coherent fictional persona may not be.

This is the harder version of Mementum's argument. The easier version, that “Noxa wrote a prompt, so he authored the output,” is unlikely to succeed under current law.

What the court will probably focus on

A few specific factual and legal questions will likely determine the outcome.

1. What did Noxa actually do, and is there evidence of it?This is the central factual question. The court will want documentation of Noxa's creative process: the prompts used, the iterations chosen, the manual edits applied (if any), the timing of the lore development, the relationship between the original AI output and the character as it now exists in commerce. Mementum's evidence here will determine whether the human-authorship argument has a foundation. Without strong contemporaneous documentation, the argument weakens.

2. Is the “character” separable from the “image”? U.S. copyright law has a substantial body of case law on character copyright, distinct from the underlying images. Mickey Mouse is a copyrightable character regardless of which specific drawing of him is being analyzed (see Walt Disney Co. v. Air Pirates, 581 F.2d 751 (9th Cir. 1978)). Sherlock Holmes is a copyrightable character described in copyrightable text. The Mementum theory might find more traction if framed as character copyright (with Noxa as author of the character's name, lore, and expression) rather than image copyright (where AI authorship issues dominate).

3. How much does the cultural specificity of the prompt matter?The Tung Tung Tung Sahur prompt isn't generic. It draws on specific Indonesian Muslim Ramadan tradition, references a specific physical artifact (the kentongan drum), and embeds a specific narrative (“visits if called for sahur three times unanswered”). The court may be asked to consider whether prompts that incorporate substantial original written expression carry different copyright weight than generic prompts.

4. The DMCA threat and procedural posture.The plaintiffs allege Mementum threatened DMCA takedowns unless paid. This is part of why Spyder Games sought declaratory relief: to obtain certainty before facing further enforcement. Whether and how Mementum sent those threats may affect the court's view of the case's ripeness and the parties' relative positions.

What the existing law actually says

A neutral read of the legal landscape, with the relevant authorities cited.

The USCO position (most relevant authority): AI-generated content alone is not copyrightable. Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16,190 (March 16, 2023). The position was elaborated in Copyright and Artificial Intelligence, Part 2: Copyrightability (January 2025), which addressed prompt-based generation specifically and concluded that prompts “do not appear to adequately determine the expressive elements produced, or control how the system translates them into an output.” This is the strongest authority for the Spyder Games position.

Thaler v. Perlmutter: The leading case. Stephen Thaler attempted to register an image authored by his AI system “Creativity Machine,” with himself as owner. The Copyright Office refused; the D.C. District Court affirmed; the D.C. Circuit affirmed (March 2025); the Supreme Court denied cert (March 2026). The human-authorship requirement is settled at the constitutional and statutory level. Thaler however addressed a case where the human did not claim authorship at all. It does not directly answer what level of human input crosses the threshold into copyrightability for AI-assisted works.

Original elements doctrine: 17 U.S.C. §103(b) provides that copyright in a derivative work or compilation extends only to the original material contributed by the author. Applied to AI works, this is the doctrinal hook for Mementum's argument: even if the AI image isn't copyrightable, original elements layered onto it (text, lore, modifications) can be protected separately.

Character copyright doctrine: Established cases include Nichols v. Universal Pictures Corp., 45 F.2d 119 (2d Cir. 1930) (the “sufficiently delineated” character test); Sam Spade v. Warner Bros., 216 F.2d 945 (9th Cir. 1954); and DC Comics v. Towle, 802 F.3d 1012 (9th Cir. 2015) (the three-part test for character protection: physically distinct, sufficiently delineated, especially distinctive). Whether Tung Tung Tung Sahur meets these criteria is a fact question.

Meme IP precedent (factually distinguishable but philosophically relevant): Keyboard Cat and Charlie Bit My Finger were not AI-generated. They were filmed videos of real subjects. The IP claims rested on copyright in the underlying video recordings, which were uncontroversially copyrightable. Mementum's reliance on these as analogies is rhetorically compelling but legally weaker than it sounds. The AI-authorship issue isn't present in those cases.

The probable timeline

Federal copyright cases of this scope rarely resolve quickly. A few markers worth knowing:

• Mementum's response or motion to dismiss: typically due within 21-60 days of service. Likely already filed or imminent as of this writing.
• Discovery: 6-18 months. Both sides will need substantial document production: Noxa's creative process records, prompts, drafts, communications with Mementum, business records.
• Summary judgment briefing: typically 12-24 months from filing.
• Trial (if the case doesn't settle or resolve on summary judgment): 18-36 months.
• Appeal (likely whoever loses): another 12-24 months at the Ninth Circuit.

Many cases of this type settle before reaching summary judgment. The strong commercial pressure (Epic's Fortnite revenue, Spyder Games's Roblox traffic, Mementum's licensing pipeline) creates incentives for a negotiated resolution. A consent decree that defines specific licensing terms without resolving the underlying copyright question is a plausible outcome.

If the court does reach the merits, the decision will probably set the most influential precedent on AI character copyright in the United States. A ruling for Spyder Games would effectively render most viral AI characters public domain, freeing platforms to use them but eliminating creator-side licensing economics. A ruling for Mementum would establish a workable theory under which AI-character licensing can be commercially viable, opening a substantial new market.

What this means for AI character creators right now

Three things, with the caveat that the law is unsettled and could move in either direction.

First, document everything. The case is going to turn on factual evidence of human authorship, and the parties with the strongest contemporaneous records will be in the best position. If you create AI characters, treat your prompts, drafts, iterations, and selection notes the way a professional artist would treat their sketch books. Timestamp them. Save them. Post the lore publicly. The cost is near zero. The upside is real.

Second, the human-authored elements of your character (the name, the backstory, the catchphrases, the mythology) are on much firmer legal ground than the AI-generated visuals. If you're considering registering copyright, the USCO will accept registrations that disclaim AI-generated material and claim only human-authored elements. This is the cleaner path. Failing to disclose AI material on a registration can lead to cancellation; disclosing properly often produces a usable, if narrower, registration.

Third, Spyder Games v. Mementumis the case to watch. Whatever you read about AI character copyright over the next 18 months, this is the case that will probably be cited as the new baseline. The smart move is to track it, understand the arguments on both sides, and adjust your own practice based on what the court eventually says, not on what either party's PR suggests.

The neutral verdict

This case is sharply contested.

Spyder Games has the cleaner fit with current Copyright Office guidance and the Thaler line of cases. If the court applies the existing AI-authorship framework strictly, AI-generated images don't get copyright, prompts don't change that, and Mementum's licensing theory has limited support.

Mementum has a more sophisticated argument than its public messaging suggests, and the original-elements doctrine plus the character-copyright doctrine give them real legal hooks. If the court is willing to look at the character holistically (as a name, plus lore, plus visual, plus narrative arc) rather than analyzing the AI image in isolation, Mementum can plausibly establish protectable authorship in some elements.

The honest answer is that both sides have credible arguments, the law is unsettled in exactly the area this case puts in play, and the outcome will depend significantly on the factual record Mementum can build around Noxa's creative process. We will likely have a clearer answer in 12-24 months. Until then, anyone telling you they know who's right about Italian Brainrot copyright is either selling something or hasn't read the briefs.

Sources

• Spyder Games LLC et al. v. Mementum Lab, 5:25-cv-10248 (N.D. Cal., filed Nov. 26, 2025)
• GamesBeat, “Who owns brainrot? Fortnite skin launch renews creator community debate over AI-generated IP” (April 2025)
• Law.com Radar coverage of Spyder Games LLC v. Mementum Lab (November 2025)
• U.S. Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16,190 (March 16, 2023)
• U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (January 2025)
• Thaler v. Perlmutter, D.C. Cir. (March 2025), cert. denied (March 2026)
• 17 U.S.C. §103(b) (original elements doctrine)
• Walt Disney Co. v. Air Pirates, 581 F.2d 751 (9th Cir. 1978)
• DC Comics v. Towle, 802 F.3d 1012 (9th Cir. 2015)
• Nichols v. Universal Pictures Corp., 45 F.2d 119 (2d Cir. 1930)
• Italian Brainrot Wiki (Miraheze) coverage of Mementum Lab
• Steal a Brainrot Wikipedia entry (commercial scale data)

CharacterClaim.ai is an automated assessment tool, not a law firm. Nothing in this post constitutes legal advice or an opinion on the merits of any pending litigation. Statements about the parties reflect public filings and reported statements as of the publication date. For decisions involving copyright registration, licensing, or enforcement, consult a licensed attorney. See our methodology page for how our assessments are derived.