A real court docket tracks where AI tool becomes author

The picture

Jason Allen sued the Register of Copyrights and the U.S. Copyright Office in the District of Colorado in September 2024 after the Office refused to register his image 'Theatre D'Opera Spatial,' produced with Midjourney. The court's own docket, case No. 1:24-cv-02665, records Allen's motion for summary judgment filed 25 August 2025, followed by the Office's response and cross-motion for summary judgment, entered 16 January 2026. No ruling has been entered on either motion as of this entry's retrieval date. This is a case still in briefing, not a decided precedent, and it should not be described as one.

What the documents show

The Office's own filing, a public court record, states the facts of Allen's process directly: he 'entered text prompts at least 624 times,' each producing a grid of options, before selecting an output he found 'acceptable,' then using Midjourney's upscaler and Photoshop to 'beautify and adjust various cosmetic details/flaws/artifacts.' The filing argues this process does not add the human authorship the Copyright Act requires. This is the government's characterisation of the record in active litigation, not a judicial finding; Allen's own August 2025 motion, not separately retrieved here in full text, is described in the same docket as arguing the opposite conclusion from the same facts.

What it is allowed to decide

A pending cross-motion decides nothing yet; the docket itself supports no class or authority claim about the artwork, only that the question is squarely before the court. The most this record establishes is a live dispute over where Provenance lies in a documented, high-iteration AI workflow with subsequent human editing, precisely the mixed-process case the Zarya of the Dawn letter and the March 2023 guidance anticipated. No authority can be held or denied here in advance of a ruling; asserting one now would be guessing.

The disclosure label

An artifact produced through hundreds of AI iterations, upscaling and manual editing, pending this kind of dispute, would carry a label reading: authorship status contested and unresolved as of 16 September 2026; Provenance disputed pending court ruling; not to be treated as either registrable or unregistrable until Allen v. Perlmutter is decided; properly asserted by the production's copyright counsel tracking the case, not assumed from either party's brief.

• How many generation, selection and editing steps separate the final asset from the first AI output, and is that count documented?
• Would this asset's authorship analysis change if the case is decided against Allen, or in his favour?
• Is anyone monitoring this docket for a ruling that would change how similar in-house workflows are described?

The docket's next entry, whenever it comes, will matter more than anything argued in the briefs summarised here.