Thirty days. Five legal developments. One unresolved question.
The U.S. copyright system hasn’t been asked to answer “what counts as human authorship” with this much legal pressure since the player piano cases of the early twentieth century. The Sahni lawsuit, filed in May 2026 in the Central District of California, is the most recent pressure point. It’s probably not the last.
The Sahni Case: What Was Filed and What It Claims
Ankit Sahni, operating through Suryast U.S. Enterprises LLC, is challenging the U.S. Copyright Office’s refusal to register “Suryast”, a composite image that combines a sunset photograph Sahni took himself with Vincent van Gogh’s *The Starry Night*, processed through an AI application called RAGHAV. The USCO rejected the application on the grounds that there was “not enough human involvement.” The Review Board issued its final rejection on December 11, 2023. Sahni filed suit in federal court in May 2026.
The legal theory is specific. According to reports about the complaint, Sahni argues that his selection of the base photograph, his layout choices, and his configuration of the RAGHAV application established baseline creative elements that qualify for copyright protection. The argument isn’t that AI authorship is protectable. It’s that Sahni’s creative decisions, before and during the AI process, constitute the kind of human authorship copyright law has always protected. The AI was the tool. Sahni, he argues, was the author.
That’s a materially different theory from Thaler v. Vidal, where a fully autonomous AI system was named as the sole inventor or author. Sahni isn’t arguing the machine made art. He’s arguing he made art using a machine, and that the Office drew the line in the wrong place.
The Human Authorship Standard: Where It Came From and What It Actually Requires
The USCO’s human authorship standard predates generative AI by decades. The Office has long held that copyright protects the expression of human creative choices, not the mechanical output of processes that happen to produce images or text. The Compendium of U.S. Copyright Office Practices makes this explicit. The question has always been: how much human choice is enough?
For fully autonomous AI output, the answer is clear and consistent. The USCO won’t register it. Courts have backed that position. But Sahni’s work sits in a different category: human photographer, human artistic direction, AI as processing layer. The Office says that’s still not enough. Sahni’s lawsuit says the Office is wrong.
The problem is that the USCO has never published a precise threshold. “Not enough human involvement” is a conclusion, not a standard. The Office has evaluated registrations case by case. That ad hoc approach has worked as long as applicants couldn’t challenge it in federal court. Sahni is doing exactly that.
The Pattern: Five AI Copyright Developments in Thirty Days
The Sahni lawsuit doesn’t exist in isolation. The hub’s thirty-day registry shows a concentrated pattern of AI copyright pressure that compliance teams should map in full:
Unanswered Questions
- What level of documented human creative decision-making, input selection, style configuration, layout, satisfies the human authorship threshold?
- Does the USCO's disclaim-and-separate approach work when human and AI contributions can't be cleanly separated?
- If Congress legislates on Cox v. Sony or digital replicas before Sahni resolves, does the statutory standard supersede the administrative one?
Who This Affects
*SCOTUS cert denial (May 16, 2026):* The Supreme Court declined to review an AI authorship challenge, leaving the lower court’s human authorship ruling intact. That decision didn’t create new law, it left existing precedent undisturbed. But it also signaled the Court isn’t ready to step in, which means the circuit courts and district courts are the active battleground.
*The Anthropic settlement (May 15, 2026):* Anthropic’s $1.5B settlement resolved the training data copyright question for that specific case without creating binding precedent. The settlement’s structure, reportedly including licensing provisions, suggests training data exposure is real and quantifiable, but the legal theory remains untested in a final judgment.
*Voice actor lawsuit (reported May 14, 2026):* A separate action targeting AI voice replication puts the digital likeness question in front of courts at the same time Perlmutter is telling Congress federal legislation on digital replicas is urgently needed.
*Perlmutter’s May 12 Senate testimony:* The Register of Copyrights disclosed to the Senate Subcommittee on Intellectual Property that the USCO has registered over 7,000 claims containing disclaimed AI-generated elements, according to Holland & Knight’s May 22 analysis of the hearing. That figure tells you the Office has operationalized AI-element registration at scale, it processes these claims, it just requires human-authored elements to be identified and AI-generated elements to be disclaimed.
*Sahni v. USCO (May 2026):* The lawsuit that asks whether the human-authored component of an AI-assisted work clears the bar the Office has been applying.
These aren’t separate stories. They’re the same question arriving from five directions simultaneously.
What the Pattern Means for AI-Assisted Creators
The practical question for any creator using AI tools isn’t “can I get copyright protection?” It’s “what do I need to document to get it?” The USCO’s practice, register the human-authored portions, disclaim the AI portions, is operational. Seven thousand claims prove it. The catch is that you have to know which portions are which, you have to disclose accurately, and the human portions have to be identifiable as genuinely creative choices, not just prompts.
Sahni is testing the harder case: what if the human and AI contributions are genuinely intertwined, not separable? His photograph and van Gogh’s style, merged by RAGHAV, aren’t easily parsed into “human here, AI there.” The USCO said that’s the problem. Sahni says that’s wrong.
Warning
The USCO has processed 7,000+ AI-element registrations under a disclaim-and-separate approach. That approach assumes human and AI contributions are separable. Sahni's case is the first federal challenge to ask what happens when they're not. That's a gap every AI content compliance program should be thinking about now, not after the ruling.
If the Central District court rules in Sahni’s favor, it will need to articulate what level of human creative decision-making clears the threshold, and that articulation will be more precise than anything the USCO has published. That precision could benefit creators. It would also constrain the Office’s discretion significantly.
If the court rules for the USCO, the existing standard gets federal validation, and the ad hoc case-by-case approach continues until Congress or the Supreme Court intervenes.
What to Watch
The docket is the first signal. When the USCO files its response, the legal theory it advances will indicate how broadly the Office intends to defend its position. A motion to dismiss would signal confidence. An answer on the merits would signal the Office expects this to go to summary judgment or trial.
Watch also for the legislative track. Perlmutter is simultaneously asking Congress to fill the *Cox v. Sony* gap and the digital replica gap. If Congress moves toward markup on any of those items, the legislative and judicial tracks start converging, and the USCO’s administrative standard could be preempted by statute before the Sahni case resolves.
State-level activity matters too. California’s AI copyright and digital likeness rules are active and enforcement-oriented. If the federal track stalls, state law fills the vacuum – which creates a compliance patchwork that’s harder to manage than a single federal standard.
TJS Synthesis
The thirty-day pattern isn’t a coincidence. AI copyright law is reaching the moment when the questions that practitioners have been asking theoretically, what counts as human authorship, what protects digital likenesses, what exposure does training data create, are arriving simultaneously in courts, in Congress, and at the USCO’s registration desk. The Sahni lawsuit is important not because it’s likely to succeed quickly, but because it forces a federal court to produce language the USCO never has. Whatever the Central District decides, it will be more precise than “not enough human involvement.” Compliance teams should treat that precision, whatever its direction, as the operative standard going forward. The USCO’s ad hoc approach has an expiration date. Federal litigation just started the clock.