Jason Allen Says His AI Art Was Copied—But His Copyright Case Is Still Pending

Jason M. Allen still has no copyright registration covering the complete version of Théâtre D’opéra Spatial, the Midjourney-assisted image he says others copied. A current Mishcon de Reya litigation tracker lists his February 27, 2026 reply as the latest key filing in the case and identifies no judgment resolving the competing summary-judgment motions.

What has changed since Allen first complained publicly about unauthorized reproductions is the procedural posture: his dispute is now a federal challenge to the registration refusal. What remains unchanged is the central legal obstacle—the effort invested in prompting does not by itself establish that Allen determined the specific expression produced by Midjourney.

The copying claims have not established ownership

Allen’s position combines two separate claims: that he authored the complete image and that third parties used it without permission. The second claim does not answer the first, because copyright infringement ordinarily requires the claimant to identify protected expression that belongs to that claimant.

In his September 26, 2024 federal complaint, Allen alleges that one person incorporated the full image into another work and that an unauthorized reproduction appeared for sale on Etsy. Those are the plaintiff’s allegations, not judicial findings of infringement, ownership or financial loss.

The lawsuit is directed at the Copyright Office and the official responsible for the challenged decision. It seeks reversal of the agency action rather than damages from the people accused of copying the picture, so the court can decide the registration dispute without determining whether any particular seller infringed Allen’s rights.

This distinction also complicates the word “stealing.” Unauthorized use may violate a copyright when protected expression is copied, but the current proceeding must first determine whether Allen can claim authorship of the generated expression in the complete image. Marketplace norms about permission and credit do not independently create copyright protection.

Why 624 prompt iterations were not enough for the Copyright Office

The Copyright Office Review Board’s September 5, 2023 decision records that Allen revised prompts at least 624 times, selected a Midjourney output, altered it in Adobe Photoshop and enlarged it with Gigapixel AI. The document also records the image’s success at the 2022 Colorado State Fair and affirms the refusal to register the work as submitted because Allen would not exclude the AI-generated material from his claim.

The number of iterations demonstrated a sustained creative process, but the agency focused on control over the resulting expression. Its analysis separated instructions about subject, composition, tone or appearance from the precise visual elements Midjourney produced when processing those instructions.

Allen’s legal theory treats the sequence as a connected act of authorship: he formed the concept, refined directions, evaluated outputs and continued until the image matched his creative objective closely enough for further editing. The competing view is that this process shows influence and selection, while leaving the system responsible for many of the particular visual choices fixed in the selected output.

That is why comparisons with photography or directing do not resolve the dispute by themselves. The relevant question is not whether creators commonly use machines or delegate tasks, but whether this creator exercised sufficient control over the protectable expressive features of this result.

The Photoshop work is a narrower copyright question

The refusal does not create a blanket rule that using generative AI makes an entire project ineligible for copyright. A composite work can contain both unprotectable generated material and protectable human expression, with copyright limited to the human-authored contribution.

The administrative history treated Allen’s later edits separately from the Midjourney foundation. Registration staff had previously viewed his claimed Photoshop visual edits as containing sufficient original authorship, while the Review Board ultimately declined to determine whether the specific adjustments were independently copyrightable because the record lacked enough information.

That difference matters for the scope of any future registration. Protection for qualifying edits would not automatically give Allen exclusive rights over every feature already present in the generated image; an infringement claim based on those edits would need to identify the human-created elements that another party copied.

Gigapixel AI occupies an even narrower place in the dispute. Allen acknowledged in the administrative record that enlarging the image did not introduce new original elements, so upscaling does not supply the missing authorship claim for the underlying composition.

The court is reviewing authorship, not the value of Allen’s effort

The case places three stages of the workflow under different legal lenses: instructions supplied to Midjourney, expression generated by the system and changes made after an output was selected. Collapsing those stages into a single label such as “AI art” would conceal the question the court must resolve.

Allen’s strongest point is not simply that he worked for many hours. It is that his repeated choices allegedly formed a sufficiently controlled creative process to make the finished expression his own. The government’s position instead treats the generated details as outside his authorship because his prompts influenced, but did not determine, how those details appeared.

A decision for Allen could establish that iterative direction of a generative model sometimes supplies enough human control for authorship. Such a result would still require courts and applicants to distinguish unusually directed workflows from routine prompting, regeneration and selection.

A decision upholding the refusal would be narrower than declaring all AI-assisted works unprotected. It could leave copyright available for original text, editing, arrangement or other identifiable human contributions while excluding generated expression that the user did not sufficiently control.

What the pending case means for creators

No court has yet awarded Allen copyright in the complete image. His allegations of copying show why the boundary has commercial consequences, but they do not establish the ownership necessary to enforce rights in every visible element.

The eventual ruling may clarify whether a long sequence of prompts can cross the line from communicating ideas to determining expression. It may also leave much of the analysis dependent on the capabilities of a particular system and the evidence showing how a creator controlled the final result.

For now, the dispute illustrates a practical difference between receiving credit for a creative process and owning enforceable rights in its output. Allen can identify himself with the project and object to unwanted reproductions, but his federal case must establish authorship before those objections become a copyright claim covering the complete work.

Also read:

• 10 Old-School Marketing Tactics that Still Work
• Chef Admits His Smash Hit Pizza Was Invented by ChatGPT
• Even People Who Hate AI Art Appear to Actually Prefer AI Art in a Blind Test