The Supreme Court gets roughly 7,000 to 8,000 cert petitions each term. It accepts around 60 to 70. When the Court declines a case, it issues no opinion and gives no reason. Silence is the answer.
On March 2, 2026, the Court gave that answer to Stephen Thaler. His petition in Thaler v. Perlmutter (Case No. 25-449) asked the justices to decide whether a work autonomously created by an AI system could qualify for copyright protection. They declined. The D.C. Circuit’s ruling affirming the Copyright Office’s human authorship requirement now stands as controlling appellate authority in that circuit. Thaler’s effort to establish AI authorship through US courts is, as Holland & Knight’s IP/Decode Blog noted, effectively at an end.
What the Court did not do is equally important. It did not define the outer boundary of “human authorship.” It did not address AI-assisted work, the category covering the vast majority of actual enterprise AI content workflows. It left open every question about what level of human creative input is sufficient to establish authorship. Those questions are alive in other circuits and at the Copyright Office.
The human authorship spectrum, where your workflow fits
Not all AI content is the same under copyright law. The framework that matters is not “was AI involved?” It’s “where did the creative judgment come from?”
Current Copyright Office guidance, consistent with the D.C. Circuit’s ruling, draws a workable distinction:
*No copyright available:* A user prompts an AI system and accepts the output as generated, with no meaningful selection, arrangement, or creative modification. The output is what DABUS produced in Thaler’s case: the AI made every creative decision. This is the category Thaler v. Perlmutter closes off.
*Copyright likely available:* A human author uses AI as a tool, selecting from outputs, arranging elements, modifying generated content, writing portions directly, in a process where the human is exercising creative judgment. The resulting work reflects human authorship in the same way a work created using Photoshop or a word processor does. The tool doesn’t displace the author.
*Contested middle ground:* Complex iterative workflows where the human provides extensive prompting direction, curates outputs across multiple generations, and composes a final work from selected elements. The Copyright Office has approved some registrations in this category and denied others. The “threshold of human authorship” in these cases is not yet a bright line.
The practical question for your organization: Do you know where your content workflows fall on this spectrum? If a copyright claim matters for a given output, can you document the human creative decisions that support it?
What protection is actually available for AI output
Copyright isn’t the only form of protection. For companies generating valuable AI output, the relevant alternatives are:
*Trade secret.* If the output is valuable and kept confidential, trade secret law may protect it regardless of copyright availability. This requires actual secrecy practices, access controls, confidentiality agreements, documented handling procedures. If your AI-generated competitive intelligence reports never leave your organization, copyright’s absence is less consequential.
*Contract.* Copyright protects against the world. Contract protects against the counterparty. If you’re licensing AI-generated content to clients, your license agreement governs use regardless of copyright status. Terms of service can restrict reproduction, redistribution, and derivative use even where copyright law wouldn’t provide a remedy against a third party.
*Database rights.* In the EU, sui generis database rights protect substantial investment in gathering, verifying, or presenting data, independent of whether individual elements are copyrightable. This is a meaningful protection layer for companies compiling AI-generated datasets or content collections, particularly for operations in EU jurisdictions.
*Trademark.* Brand elements associated with AI-generated content, names, logos, trade dress, remain protectable through trademark regardless of underlying content copyright status.
What Congress would need to do
Legal analysts note that any expansion of US copyright to accommodate AI authorship would require Congressional action. This is the direct consequence of the cert denial: the judicial path is closed, at least through this set of facts.
What would Congress need to address? At minimum: whether copyright protection attaches to AI output and, if so, who holds it (the developer of the AI system, the user who prompted it, the company that deployed it). These aren’t small drafting questions. They implicate the entire incentive structure copyright is designed to create, and there’s no consensus in Congress, in the academic literature, or in the tech industry about the right answer.
Near-term Congressional action on AI copyright is not well-supported by current legislative activity. Flag it as a longer-horizon risk for IP strategy planning, not an imminent development.
How other jurisdictions compare
The US position isn’t unique, but the rationale varies.
In the UK, the Copyright, Designs and Patents Act includes a specific provision for computer-generated works, assigning authorship to “the person by whom the arrangements necessary for the creation of the work are undertaken.” Whether that covers modern AI-generated content under current interpretations is contested, and the UK government has been reviewing the provision. The UK position is more receptive to AI authorship in principle than the US, but the law’s practical scope for current generative AI outputs isn’t settled.
In the EU, the EU AI Act’s transparency obligations (effective August 2, 2026) focus on labeling AI-generated content rather than its copyright status. The EU’s approach to AI copyright sits primarily in the Copyright Directive framework and ongoing policy discussions, not in the AI Act itself.
Neither the UK nor EU position changes the US compliance calculus for companies primarily concerned with US copyright. They do matter for companies with IP strategies spanning multiple jurisdictions.
The practical checklist
Three things worth doing now, not because the cert denial changed anything, but because it confirmed that the current framework isn’t changing through the courts:
1. Audit your content workflows. Identify which workflows produce AI-assisted content (human authorship likely) versus AI-generated content (human authorship absent or unclear). The distinction drives your copyright strategy.
2. Document human creative decisions. For workflows where copyright matters, maintain records of the human selections, modifications, and creative judgments involved. Copyright Office registration applications for AI-assisted works benefit from this documentation.
3. Don’t assume copyright where it doesn’t exist. If your organization is relying on copyright protection for AI-generated marketing copy, product descriptions, or other content, verify that assumption against your actual workflows. If the output is fully autonomous, the protection isn’t there.
The Supreme Court’s silence in Thaler v. Perlmutter is the clearest signal the courts have sent on this issue. Take it at face value.