The D.C. Circuit ruled that AI-only works receive no copyright protection.

The D.C. Circuit ruled that AI-only works receive no copyright protection. That's the current law, and it creates a massive IP ownership gap for firms and businesses using AI-generated images in marketing, presentations, and client materials. If a human didn't make creative choices that shaped the output, the output belongs to everyone -- or no one.

Prompt engineering doesn't clearly bridge that gap. Some IP scholars argue that sufficiently detailed prompts constitute human authorship. The Copyright Office hasn't agreed yet. Until there's definitive guidance, every AI-generated image your firm uses carries an ownership question mark that your clients' competitors can exploit.

The D.C. Circuit Ruling: AI-Only Works and Copyright

In Thaler v. Perlmutter, the D.C. Circuit upheld the Copyright Office's refusal to register a work created autonomously by an AI system (DABUS) without human creative input. The court held that copyright requires human authorship -- a principle rooted in the Constitutional text granting Congress power to secure rights to "Authors."

The ruling was narrow: it addressed works created entirely by AI without human creative involvement. It explicitly left open the question of works where humans use AI as a tool -- providing detailed prompts, selecting among outputs, making post-generation edits. That open question is where the current uncertainty lives.

For IP lawyers advising clients, the safe ground is clear: if a human made expressive creative choices that shaped the final work, there's a copyright argument. If the human's contribution was limited to pressing "generate" with a basic prompt, there isn't. Everything in between is unsettled law.

What Prompt Engineering Adds (or Doesn't) to Copyright Claims

The Copyright Office's guidance on AI-generated works (March 2023, updated 2025) draws a line between mechanical and creative human contributions. Selecting an AI tool and entering a prompt is mechanical. Crafting a highly detailed prompt that specifies composition, style, color palette, subject positioning, and artistic references is... still uncertain.

The Office has registered works where AI-generated elements were substantially modified by human post-generation editing. It has rejected works where the human contribution was limited to prompt writing, even detailed prompts. The implicit position: prompts are instructions, not expressions.

This creates a practical problem for businesses generating AI images at scale. If prompt engineering alone doesn't establish authorship, every AI-generated marketing image, every AI-created illustration, every AI-designed logo exists in a copyright vacuum. Competitors can copy them freely. Clients can't enforce exclusivity.

The IP Ownership Gap for AI-Generated Visual Content

The ownership gap has real business consequences. A law firm that uses AI to generate images for its website can't stop a competitor from using identical images -- because neither firm owns them. A client who commissions AI-generated product photography can't prevent counterfeitors from using the same images.

Trade secret protection doesn't help because AI-generated images are published. Trademark protection can apply to AI-generated logos that acquire secondary meaning, but the underlying image remains uncopyrightable. Contract law can restrict use between parties (license agreements, terms of service) but can't create rights against the world.

For IP lawyers, the advisory obligation is clear: clients using AI-generated visuals in business-critical applications need to understand the ownership limitations. Relying on AI-generated images for brand identity, product differentiation, or competitive positioning carries risks that traditionally created images don't.

How Different AI Image Tools Affect the Analysis

Not all AI image generation is equal from an IP perspective. Tools like Claude Design that produce structured visual outputs (diagrams, charts, slides) aren't generating "artistic works" in the traditional sense -- they're closer to automated formatting tools, and the content they format may have separate copyright protection.

Tools like Midjourney, DALL-E, and Stable Diffusion generate novel images from text prompts. These outputs face the full weight of the authorship question. The more the tool makes autonomous creative decisions (composition, color, style, detail), the weaker the human authorship argument.

Tools that enable significant human post-generation editing (inpainting, outpainting, style transfer on human-created base images) produce works with stronger copyright claims because the human creative contribution is more substantial and identifiable.

For clients choosing AI image tools, the IP analysis should inform the selection. Use fully generative tools for disposable content (social media posts, internal presentations). Use human-in-the-loop tools for brand assets and materials requiring IP protection.

Practical Guidance for Law Firms and Their Clients

Document human creative contributions meticulously. If an employee writes a detailed prompt, selects among 50 generated options, edits the chosen image in Photoshop, and makes additional creative decisions -- document each step. That documentation is your copyright registration evidence.

Don't rely on AI-generated images for trade dress or brand identity without understanding the ownership gap. A competitor who recreates your AI-generated visual identity can't be stopped with copyright alone.

Consider hybrid workflows: use AI for initial concepts and rough compositions, then have human designers execute the final versions. The human-created final work has clear copyright protection while still capturing the speed advantages of AI ideation.

Stay current on Copyright Office guidance. This area is evolving rapidly, with new guidance expected in late 2026 that may address the prompt engineering question directly. Advise clients based on current law, but design workflows that can adapt to favorable changes.

The Bottom Line: AI-only images have no copyright protection under current law, and prompt engineering alone probably doesn't change that -- IP lawyers need to advise clients on the ownership gap before it becomes a business problem.