Skip to content
BETA

AI-generated and AI-assisted images

Current as of Tuesday, September 1, 2026. General information, not legal advice.

RightsValet Copyright Guide: all sections

Ready to register your photos? Request an invitation

19.1 Is the March 16, 2023 Registration Guidance still in force?

Yes, as the Office’s posted registration policy. It is a statement of policy, effective March 16, 2023, 88 FR 16190 (Mar. 16, 2023), FR Doc. 2023-05321, 37 CFR Part 202. Live hub https://www.copyright.gov/ai/ still links the PDF. Part 2 (Jan. 29, 2025) affirmed the Guidance’s human-authorship framework and stated the Office “plans to supplement its March 2023 registration guidance and update the relevant sections of the Compendium.” As of the 2026-09-01 fetch of copyright.gov/ai, no replacement circular or superseding FR registration policy was posted. Treat 88 FR 16190 as current registration practice unless and until the Office issues the promised supplement.

Federal Register: https://www.federalregister.gov/documents/2023/03/16/2023-05321/copyright-registration-guidance-works-containing-material-generated-by-artificial-intelligence PDF: https://copyright.gov/ai/ai_policy_guidance.pdf

Human-authorship baseline in the Compendium (Third) is unchanged: § 306 - “The U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being.” § 313.2 - the Office “will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.” Circular 1: “independently created by a human author.” Compendium § 905: product of human authorship.

19.2 Subsequent Office AI reports (Parts 1–3) and registration-practice status

Part Date Topic Effect on still-image registration
Part 1 July 31, 2024 Digital replicas (voice/appearance) Recommends federal legislation; not a registration circular for photographs. https://copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf
Part 2 January 29, 2025 Copyrightability of generative-AI outputs The operative Office analysis for AI-assisted stills. Concludes existing law is adequate; no new copyright or sui generis right for purely AI-generated material. Promises additional registration guidance + Compendium update (not yet posted as of 2026-09-01). https://copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf
Part 3 Pre-publication May 9, 2025 Generative AI training, fair use, licensing Explicitly not a registration-practice document. Final version “without any substantive changes expected.” https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf

NOI: 88 FR 59942 (Aug. 30, 2023).

Part 2 executive conclusions (Office language): questions of copyrightability and AI can be resolved pursuant to existing law; the use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output; copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material; copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements; based on the functioning of current generally available technology, prompts do not alone provide sufficient control; human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs.

19.3 Office-published registration decisions (still images / visual works)

Cite only these Office documents (not later court dockets). The Review Board opinions index (https://copyright.gov/rulings-filings/review-board/), as fetched 2026-09-01, lists no later “human authorship / AI-generated” still-image decisions after SURYAST. Later 2024–2026 Board entries on that page are logos, useful articles, and similar refusals, not AI-authorship still-image cases.

1. A Recent Entrance to Paradise - Review Board, Feb. 14, 2022. https://www.copyright.gov/rulings-filings/review-board/docs/a-recent-entrance-to-paradise.pdf . Work described as “autonomously created by a computer algorithm running on a machine.” Refusal affirmed: no human authorship. The Board also rejected a WFH theory that a machine could be an “employee.” 88 FR 16190 and Part 2 both rest on this decision. Part 2 (Jan. 2025) notes the D.D.C. affirmed the Office in Thaler v. Perlmutter, 687 F. Supp. 3d 140 (D.D.C. 2023), and that “[t]he decision has been appealed.” This report does not report later appellate or Supreme Court outcomes; they are not Office-published.

2. Zarya of the Dawn - Cancellation / reissuance letter, Feb. 21, 2023 (Robert J. Kasunic). https://www.copyright.gov/docs/zarya-of-the-dawn.pdf . Registration VAu001480196 cancelled and replaced. Human-authored text and human selection, coordination, and arrangement of text and images are protected. Individual Midjourney-generated images are not. On prompts: “A person who provides text prompts to Midjourney does not ‘actually form’ the generated images and is not the ‘master mind’ behind them. … prompts function closer to suggestions than orders, similar to the situation of a client who hires an artist to create an image with general directions as to its contents.” Minor Photoshop edits to a mouth were “too minor and imperceptible.” Sweat of the brow (hundreds of iterations) is irrelevant (citing Feist and Compendium § 310.7).

3. Théâtre D’opéra Spatial - Review Board, Sept. 5, 2023. https://copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf . Jason Allen used Midjourney (~624 prompt revisions), then Photoshop, then Gigapixel AI upscaling. Refusal affirmed because he would not disclaim more-than-de-minimis AI content. Prompting, however elaborate, does not make the user the author of the Midjourney image. On upscaling (applicant concession, accepted by the Board): Gigapixel AI “doesn’t introduce new, original elements into the image” and “the enlargement process … does not equate to authorship.” On Photoshop / content-aware fill: Adobe’s content fill “fills empty spaces … with little or no input from a user, which suggests a lack of human authorship of filled material”; a newer version employs machine learning “to actually create new content for an image.” Fair use is not a copyrightability doctrine. Applicants need not list each AI tool or percentages - a “brief statement” such as “generated by artificial intelligence” suffices.

4. SURYAST - Review Board, Dec. 11, 2023. https://copyright.gov/rulings-filings/review-board/docs/SURYAST.pdf . Photographer Ankit Sahni input his own photograph + van Gogh’s The Starry Night + a style-transfer strength value into RAGHAV. The Board affirmed refusal. RAGHAV, not Sahni, determined placement, colors, and whether elements appeared. Selecting a filter-strength number is de minimis (citing Compendium § 909.3(A)). The original photograph remains a separate, potentially registrable work; the AI-modified output is not. “Unlike Rose Enigma, the output did not clearly show the copyrightable work input by the applicant” (Part 2’s later summary).

5. Rose Enigma - described in Part 2 (not a Review Board opinion): VAu001528922 (Mar. 21, 2023). Human hand-drawn illustration used as input; AI output still showed the drawing. Registered with annotation: “Registration limited to unaltered human pictorial authorship that is clearly perceptible in the deposit and separable from the non-human expression that is excluded from the claim.”

19.4 What must be disclosed (limitation of claim / disclaim AI)

88 FR 16193 (Guidance § IV.A): “applicants have a duty to disclose the inclusion of AI-generated content in a work submitted for registration and to provide a brief explanation of the human author’s contributions to the work.”

Mechanics:

  • Use the Standard Application (Single Application cannot be used if not all content is by the same individual).
  • Author Created: describe human authorship only. Example for a compilation: “Selection, coordination, and arrangement of [human content] created by the author and [AI content] generated by artificial intelligence.”
  • Do not list the AI system or vendor as author or co-author.
  • Limitation of Claim → Material Excluded → Other: “[description of content] generated by artificial intelligence.” Required for AI content that is more than de minimis. De minimis (brief quotes / trivial amounts that would not themselves be copyrightable) need not be disclaimed (Compendium § 503.5, cited in Guidance n.40).
  • Optional: Note to CO. If unsure, a general statement that the work contains AI-generated material is enough; the Office will correspond. “In some cases, the use of an AI tool will not raise questions about human authorship, and the Office will explain that nothing needs to be disclaimed.”
  • January 27, 2024 eCO update: onscreen guidance includes “material generated by artificial intelligence.”

Pending applications: contact PIO to add a note. Already-issued registrations: supplementary registration (Circ. 8; $100 electronic). Failure to disclose can lead to cancellation (37 C.F.R. § 201.7(c)(4)) or § 411(b) inaccuracy problems. June 28, 2023 Office webinar transcript: https://copyright.gov/events/ai-application-process/Registration-of-Works-with-AI-Transcript.pdf - “appreciable” ≈ more than de minimis; would the AI material support a derivative-work claim if a human had added it?

88 FR 16192: “If a work’s traditional elements of authorship were produced by a machine, the work lacks human authorship and the Office will not register it. … when an AI technology receives solely a prompt from a human and produces complex written, visual, or musical works in response, the ‘traditional elements of authorship’ are determined and executed by the technology - not the human user. … these prompts function more like instructions to a commissioned artist.” A human may select or arrange AI-generated material in a sufficiently creative way that “the resulting work as a whole constitutes an original work of authorship.” Or an artist may modify material originally generated by AI technology to such a degree that the modifications meet the standard for copyright protection. Copyright will only protect the human-authored aspects.

Assistive tools vs. generative systems (88 FR 16192–93): “a visual artist who uses Adobe Photoshop to edit an image remains the author of the modified image … what matters is the extent to which the human had creative control over the work’s expression and ‘actually formed’ the traditional elements of authorship.” Footnote 36: “To the extent, however, that an artist uses the AI-powered features in Photoshop, the edits will be subject to the above analysis.” Part 2 § II.C (assistive uses the Office agreed should not destroy copyright in the human work): “removing unwanted objects or crowds from a scene”; “color correction, detail sharpening, or de-blurring.” “This distinction depends on how the system is being used, not on its inherent characteristics.” Part 2 on prompts: iterative re-prompting is still “re-rolling the dice.” Authorship-by-adoption (picking a liked output) is rejected. Compilations of only two or three elements are generally de minimis (Compendium § 312.2).

19.5 Can you register a photograph run through AI upscaling, sky replacement, or generative fill?

Cite Office language only. The Office has not published a decision that uses the words “sky replacement” or “generative fill” as product names. The following is what the Office has said that maps onto those techniques.

Upscaling: Théâtre: enlargement via Gigapixel AI “does not equate to authorship.” Compendium § 313.2 examples of machine processes lacking human authorship include “Reducing or enlarging the size of a preexisting work of authorship” and converting analog to digital. Compendium § 909.3(A): typical technical alterations (sharpening and balancing colors, tint, tone) do not warrant registration as new authorship. Part 2: assistive “detail sharpening” / “de-blurring” does not bar copyright in the underlying human photograph. Office-consistent reading: a human-authored photograph that is only upscaled, sharpened, or denoised remains registrable as that photograph. The upscaling step itself is not new copyrightable authorship. No AI disclaimer is required for purely technical enlargement if it does not insert new expressive AI content. If the upscaler hallucinates new detail (new textures, objects, faces), that new material is AI-generated and, if more than de minimis, must be disclaimed.

Generative fill / content-aware fill / inpainting: Guidance n.36: AI-powered Photoshop features are subject to the generative-AI analysis. Théâtre: content-aware fill “suggests a lack of human authorship of filled material.” Part 2: “removing unwanted objects or crowds from a scene” was listed among assistive uses that “should not affect” copyright in the human work. That is object removal; the Office did not say the filled-in pixels are copyrightable. If fill generates new expressive content (a new sky, a new wall, a new sofa) determined by the model, 88 FR 16192 requires that material to be disclaimed if more than de minimis.

Sky replacement: Not named in 88 FR 16190, Part 2, Zarya, Théâtre, or SURYAST. Closest Office analogs: (1) SURYAST - an AI process that replaces the expressive treatment of a photograph’s sky/scene under a style image was not human authorship of the output; the original photo could still be registered separately. (2) Guidance: if “the AI technology determines the expressive elements of its output, the generated material is not the product of human authorship … and must be disclaimed.” A photographer who photographed a sky and compositing it by hand (traditional composite) is ordinary derivative/photographic authorship under Compendium ch. 900. A photographer who prompts an AI to invent a sky is in the Zarya / prompt-to-image column: the new sky is AI-generated and, if more than de minimis, must be excluded. The human photograph of the house/land, if still perceptible, remains claimable (Rose Enigma logic).

Fully generated images (text-to-image with no human photograph): Zarya, Théâtre, Paradise, Part 2 - not registrable as to the image itself. Selection/arrangement of many such images plus human text can support a compilation claim (Zarya).

Application practice for an AI-edited photograph: Standard Application; Author Created = “photograph”; Material Excluded = e.g. “sky and filled regions generated by artificial intelligence” (if those regions are more than de minimis); New Material Included = “photograph” or “human-authored photographic authorship other than AI-generated sky/fill.” Do not name Midjourney/Photoshop/Firefly as author.