US Copyright Office AI Policy 2026: What Businesses Need to Know

The Core Copyright Office Position on AI-Generated Content

The Copyright Office's position, consistent since 2023 and reinforced in its 2025 final report, rests on a single principle: copyright requires human authorship. The Office will not register works "produced by a machine or mere mechanical process that operates without any creative input or intervention from a human author."

Applied to AI, this means that purely AI-generated content — where a human types a prompt and the AI generates the work — is not protectable. The human author didn't make the specific expressive choices that copyright law rewards. The AI did.

But the analysis doesn't end there. Most business use of AI isn't purely hands-off. Human creative choices made in the process of using AI — choices about which AI outputs to select, how to arrange them, how to modify them, what to combine them with — can themselves be original expression that copyright protects. The Office will register those human contributions while excluding the AI-generated material.

What This Means for Your Content Team

The gap between "publishable AI content" and "copyrightable AI content" is where most businesses have blind spots. Your marketing team can publish AI-generated blog posts. Legal challenge arises when a competitor repurposes that content and you want to enforce. If the published content is purely AI-generated, you may have no copyright claim to enforce.

Three content scenarios and their copyright status:

Copyright Registration in 2026: The AI Disclosure Requirement

Since March 2023, the Copyright Office requires applicants to disclose AI-generated content when submitting registration applications. The process:

• New applications: Use the "Nature of Authorship" field to describe which elements are human-authored and which were AI-generated. Example: "Text authored by [name], excluding AI-generated passages; selection and arrangement of content."
• Existing registrations: If you registered AI-involved works before the disclosure requirement and didn't disclose AI involvement, you may need to file a supplementary registration to correct the record. Registrations with material omissions can be challenged.
• What happens after disclosure: The Office examines the human-authored elements and issues registration for those elements only. The AI-generated portions receive no protection.

Training Data: Where the Legal Risk Is Highest for AI Companies

If your company trains AI models — including fine-tuning foundation models on your proprietary data — training data copyright is where you face the most significant business risk. Multiple lawsuits filed against major AI companies since 2022 are still working through the courts as of 2026.

The Copyright Office's 2025 report on training data declined to declare AI training on copyrighted content fair use as a categorical matter. Instead, it identified the key fair use factors courts will weigh:

• Purpose and character of the use: Commercial training on web-scraped content is more vulnerable than research use; transformative processing is more protected than reproduction.
• Nature of the copyrighted works: Creative works (fiction, music, visual art) receive stronger protection than factual works.
• Amount used: Training on entire works is more problematic than training on excerpts.
• Market harm: Whether the AI output substitutes for or competes with the original market is increasingly central to court analysis.

The practical risk profile: companies training on licensed data have much stronger legal footing than companies training on scraped data without licenses. The Office explicitly noted that the market for licensed training data is developing and recommended Congress consider licensing frameworks.

The Output Infringement Question

Separate from training data liability is the question of whether AI-generated outputs infringe the training data. Courts have begun addressing this in specific cases with mixed results. Key factors that increase output infringement risk:

• AI generates content that closely reproduces recognizable portions of training works
• Users deliberately prompt AI to reproduce copyrighted content ("write this in the style of X" or "reproduce the lyrics to...")
• AI is trained on a narrow corpus where specific works are heavily weighted
• Outputs are used commercially in ways that compete with the original market

If your business uses a third-party AI platform and an AI-generated output turns out to infringe a third party's copyright, your exposure depends significantly on your vendor contracts. Review your AI platform's terms of service for indemnification provisions — some major providers offer IP indemnification for business customers, while others disclaim all liability.

The Compliance Checklist for AI-Using Businesses

What's Coming: Legislative Watch

Congress has held multiple hearings on AI copyright since 2023 but has not passed legislation as of mid-2026. The Copyright Office recommended Congress consider: a licensing framework for AI training on copyrighted works; transparency requirements for AI-generated content disclosure; and clarification of authorship standards for AI-involved works. Watch for movement in these areas — legislative changes could significantly alter the risk calculus for businesses built on AI-generated content or trained models.

Several bills have been introduced addressing AI and copyright, including proposals for mandatory training data disclosure, licensing requirements, and creator compensation funds. None have passed but the legislative environment is active. Subscribe to Copyright Office news updates and work with IP counsel to stay current.