New Lawsuit Challenges Copyright Office's Refusal to Register AI-Assisted Artwork 'Suryast'

A new federal lawsuit filed on May 8, 2026 in the U.S. District Court for the Central District of California takes direct aim at one of the most contested questions in AI law: can a work created with meaningful human direction but substantial machine contribution ever be registered with the U.S. Copyright Office? Case number 2:26-cv-04999, captioned Suryast U.S. Enterprises, LLC v. Perlmutter, asks a federal judge to overturn the Copyright Office's refusal to register an AI-assisted painting titled Suryast and, more broadly, to narrow the Office's "human authorship" test that has frozen thousands of generative-AI works out of the federal registry.

The complaint, filed by attorney Ryan Abbott of Brown Neri Smith & Khan LLP, names Register of Copyrights Shira Perlmutter and the U.S. Copyright Office as defendants. Abbott is not a random lawyer. He is the same counsel who drove the multi-year DABUS campaign to secure patents and copyrights for AI systems around the world, and he represented Stephen Thaler in the long-running Thaler v. Perlmutter litigation over the AI-generated image A Recent Entrance to Paradise.

What the Suryast work is

Suryast is a digital painting first created in 2020 by Indian lawyer and artist Ankit Sahni using an AI tool called RAGHAV. Sahni provided a base photograph he had taken, chose a Van Gogh–style reference image, and set a "variable value" that controlled how strongly the AI applied the style transfer. RAGHAV then produced the final output. The work became a cause célèbre in 2021 when the Indian Copyright Office initially registered it listing both Sahni and the AI as co-authors, a registration Indian authorities later moved to withdraw after the ruling drew international attention.

The U.S. Copyright Office, by contrast, has consistently refused to register Suryast. In a 2023 decision the Review Board found that the image's "expressive elements of authorship" were not produced by a human and therefore fell outside the scope of copyrightable subject matter under Section 102(a) of the Copyright Act and the Office's 2023 registration guidance on AI-generated material. The new complaint attached to the docket runs 64 pages, including a civil cover sheet, and squarely targets that refusal.

The legal theory

According to the docket entry and public filing, Suryast U.S. Enterprises is bringing the action as a Section 411(a) appeal from the final refusal, paired with Administrative Procedure Act claims. That combination mirrors the posture Abbott used in Thaler, where the D.C. Circuit ultimately affirmed the Office's position in March 2025 that a work "autonomously created" by a machine cannot be registered. Suryast is designed to test a materially different fact pattern.

Three distinctions matter. First, Sahni supplied the input photograph, which is itself a copyrightable work. Second, he selected the style reference and tuned a numerical parameter that directly shaped the output. Third, he curated and accepted the final image as his creative expression. The plaintiff's argument, based on the relief sought, is that this level of human creative control is indistinguishable in kind from the guided use of Photoshop filters, algorithmic photography, or the selection-and-arrangement creativity the Supreme Court recognized in Feist. On that theory, the Office's blanket refusal exceeds its statutory authority and constitutes arbitrary and capricious agency action under 5 U.S.C. § 706.

Why this filing matters

The Copyright Office has processed a growing backlog of AI-assisted registrations since issuing its March 2023 guidance and its three-part Copyright and Artificial Intelligence report series in 2024 and 2025. The Office's working rule is that a human must contribute the "traditional elements of authorship" and that prompts alone, no matter how detailed, are generally insufficient. Applicants have been asked to disclaim AI-generated portions of their works as a condition of registration.

Suryast is arguably the strongest vehicle yet to test that rule in court. Unlike the Thaler case, where the applicant affirmatively claimed the AI was the author, Sahni has always claimed human authorship and argued that his photographic input and parametric controls are the dominant creative contribution. If the Central District of California sides with the plaintiff, the decision would create a circuit split with the D.C. Circuit's Thaler opinion on the degree of human involvement required. If the Office's position is upheld, registrants will have near-binding guidance that even substantial prompting, parameter tuning, and source-image selection do not clear the human-authorship bar.

Either outcome would reshape registration practice for the hundreds of generative-AI applications the Office is processing each month and would influence infringement suits where defendants argue that AI-assisted works are unregistrable and therefore cannot support statutory damages or attorney's fees under Section 412.

What to watch

A few things will signal how quickly this case moves. The Office typically responds to Section 411(a) appeals with a motion to dismiss on the administrative record rather than engaging in discovery, so the first substantive briefing is likely to arrive within 60 to 90 days. The choice of venue is also notable. The Central District of California sits within the Ninth Circuit, whose copyright jurisprudence has been more protective of thin creative contributions than the D.C. Circuit, and any appeal would bypass the court that decided Thaler.

Industry watchers should also note who is not in this case. There is no AI developer defendant, no training-data claim, and no fair-use question. Suryast is a pure authorship case, which makes it an unusually clean vehicle for the appellate question Abbott has been trying to tee up for years: how much human direction is enough?

The bigger picture

With Suryast now pending, the three live fronts of AI copyright law in the United States are clearer than they have been at any point in the past year. The training-data cases against OpenAI, Anthropic, Meta, and NVIDIA continue to grind through discovery and summary judgment. The output-infringement cases, including the Midjourney matter, are testing whether generated images can themselves be derivative works. And now the registration front, long considered settled after Thaler, has a new test case with a sympathetic human plaintiff and an experienced appellate advocate.

AICopyrightLegal.com will monitor the docket and publish analysis as the Copyright Office files its response. In the meantime, creators using AI tools should continue to follow the Office's 2023 guidance on disclosing AI-generated portions of their works when registering, and document the specific human creative choices that went into any submission. Those records will matter whether Suryast succeeds or not.

Case: Suryast U.S. Enterprises, LLC v. Perlmutter, No. 2:26-cv-04999 (C.D. Cal. filed May 8, 2026). Docket information via CourtListener/RECAP.