Who Owns AI-Generated Content? Copyright, Intellectual Property, and Legal Risks for Businesses in 2026

The Short Answer

Under current U.S. law (as of mid-2026), purely AI-generated content — text, images, music, or code created entirely by AI without meaningful human creative input — cannot be copyrighted. The U.S. Copyright Office's position, affirmed in multiple rulings, is that copyright requires human authorship. AI is a tool, not an author.

However, AI-assisted content can be copyrighted when a human makes sufficient creative contributions — selecting, arranging, editing, or modifying AI output in ways that reflect human creative choices. The key legal test: did a human exercise sufficient creative control over the final work such that the work is 'essentially authored by a human'?

For most business uses of AI (drafting and heavily editing text, using AI-generated images as components in larger human-designed works, generating code that a human reviews, modifies, and integrates), the copyright situation is manageable — protectable through human creative contribution and editing. But there are real risks: AI-generated content may inadvertently reproduce copyrighted material from training data, AI tools' terms of service may claim rights you didn't expect, and the law is actively evolving in ways that could change the landscape.

This guide covers what every business should know and do right now to protect their content and minimize IP risk.

What the U.S. Copyright Office Has Actually Said

The Copyright Office has been systematically addressing AI and copyright since 2023. Key positions as of 2026:

AI-generated works without human authorship are not copyrightable. If you type 'create an image of a sunset over a mountain' into an AI image generator and use the output as-is, you do not own a copyright in that image. Anyone can copy, modify, or use it. This has been upheld in multiple Copyright Office Review Board decisions.

Works containing AI-generated material may be copyrightable if the human contribution is sufficient. If you use AI-generated elements as raw material that you substantially modify, arrange, or incorporate into a larger human-authored work, the human-authored portions (and potentially the work as a whole, excluding the purely AI-generated elements) may be protected. The Copyright Office's test: 'whether the AI-generated contributions are the result of mechanical reproduction' (no copyright) or whether a human author's 'own original mental conception' shaped the work (copyrightable).

The 'selection, coordination, and arrangement' doctrine applies. Even if individual AI-generated elements aren't copyrightable, the way you select, coordinate, and arrange them may be. Example: a book whose individual AI-generated illustrations may not be protected, but whose human-authored text, creative arrangement of illustrations, and overall structure is protected as a compilation.

Applications must disclose AI use. The Copyright Office requires applicants to disclose whether AI was used in creating the work and to describe the human author's contribution. Failure to disclose can invalidate the registration.

The rules are still being made. The Copyright Office issued formal guidance and has been conducting a multi-year AI initiative. Additional legislation is possible. The U.S. Patent and Trademark Office (USPTO) has parallel proceedings on AI and IP. Courts are hearing the first wave of AI copyright cases, and appellate rulings will shape the landscape further.

What This Means for Different Types of Content

Text content (articles, emails, reports, books): Heavily AI-assisted but human-edited text is likely protectable to the extent of the human contribution. The key is documentation: maintain drafts showing the AI output and your edits. The more substantial your editorial contribution — restructuring, fact-checking, adding original analysis, rewriting for voice — the stronger your copyright claim. Text used substantially as the AI produced it, with only light editing for grammar, is in the grey zone.

Images (AI-generated art, illustrations, design elements): Purely AI-generated images without substantial human modification are not copyrightable under current guidance. If you use AI-generated images as components in a larger design where a human makes creative decisions about composition, modification, and integration, the overall work may be protectable as a compilation or derivative work. For commercial use where copyright protection matters: AI-generated images are raw material. Substantially modify, combine, or use them in ways that involve meaningful human creative choices.

Code (AI-generated software): The same principles apply. AI-generated code that a human reviews, tests, modifies, and integrates as part of a larger software project — the project as a whole is protectable, and the human's modifications and integration decisions are authorial contributions. Code used verbatim as AI produced it: grey zone, leaning toward not protectable. Important additional considerations: AI-generated code may reproduce open-source code from training data, potentially creating open-source license compliance obligations. Tools like GitHub Copilot have faced lawsuits over this issue.

Video and audio: AI-generated voice narration, AI-generated video clips, and AI-generated music follow the same human-authorship framework. Deeply concerning area: AI voice cloning and likeness generation may implicate right of publicity laws (separate from copyright) that vary by state.

The IP Infringement Risk: Could AI Output Infringe Someone Else's Copyright?

This is the risk that keeps lawyers up at night, and it's genuinely unresolved:

The problem: AI models are trained on massive datasets that include copyrighted material. There is evidence that some models can, under certain conditions, reproduce content substantially similar to their training data. If your AI tool produces output that is substantially similar to a copyrighted work, and you publish or sell that output, you could face copyright infringement claims.

The current legal landscape: Multiple major lawsuits are working through courts (including cases against OpenAI, Microsoft, Stability AI, and others). Key questions being litigated: Is training AI on copyrighted works fair use? Are AI companies liable when their tools produce infringing outputs? Are users liable when they publish AI outputs that infringe?

What to do right now: Use AI tools that offer IP indemnification for business/enterprise users (several major platforms now provide this — check your tool's terms). Avoid using AI to create content that mimics a specific artist, author, or copyrighted work (this increases infringement risk substantially). For commercially important content, conduct reverse image searches (for visual content) and plagiarism checks (for text) before publication. Document your creative process to establish your good-faith, independent creation. And maintain appropriate insurance coverage for IP claims if content is central to your business.

The Terms of Service Trap: What You Agreed to Without Reading

Most people click 'Accept' on AI tool terms without reading them. Some contain provisions that affect your IP rights:

Ownership of outputs: Most major AI platforms (OpenAI, Anthropic, Google) assign ownership of outputs to the user — but check. Some smaller or specialized AI tools claim broader rights. Read the 'Ownership of Content' or 'Your Content' section of the terms.

License to the provider: Some terms grant the AI provider a license to use your inputs and outputs to improve their service. This is distinct from ownership — you still own the output, but the provider may have rights to use it. For business/enterprise tiers, this license is typically excluded, but verify.

Indemnification: Some business/enterprise AI tiers include IP indemnification — the provider will defend you against copyright infringement claims related to your use of their tool. This is increasingly common among major platforms and is a meaningful reason to use business tiers for content you depend on commercially.

Practical IP Protection Steps for Businesses Using AI

1. Always add meaningful human creative contribution. This is your strongest protection. Edit, restructure, fact-check, add original analysis, combine AI output with human-created content. Document your process.

1. Keep your drafts. If you ever need to establish what was AI-generated vs human-created, your editing history and draft versions are evidence. This is tedious but valuable for content where IP ownership matters commercially.

1. Disclose AI use in copyright applications. The Copyright Office requires it, and failure to disclose can invalidate registration and potentially constitute fraud on the Copyright Office.

1. Check your AI tools' terms regarding output ownership, provider licenses, and indemnification. This takes 10 minutes per tool and is the highest-ROI legal diligence you can do.

1. Avoid using AI to create content 'in the style of' specific living artists, authors, or brands. This increases both copyright and right of publicity risk.

1. For commercially critical content, screen AI outputs for potential infringement. Reverse image search for visual content. Plagiarism checkers for text. Code audit tools for AI-generated software.

1. Add an AI use disclosure to your content publishing workflow. Not necessarily public-facing disclosure — internal documentation of what AI was used for and how humans contributed. This creates a record that supports your copyright claims if they're ever challenged.

1. Watch this space. The law is changing. Major court rulings, Copyright Office regulations, and potentially federal legislation will reshape the landscape in the next 1-3 years. Put a quarterly reminder to check for significant legal developments.