U.S. Copyright Office Issues Two Reports About Artificial Intelligence

The U.S. Copyright Office is charged by statute with administering the United States’ copyright laws. It is housed in the Library of Congress, and its mission is “to promote creativity and free expression by administering the nation’s copyright laws and by providing impartial, expert advice on copyright law and policy for the benefit of all.”

One of the leading headlines since the inception of large language models (LLMs) like ChatGPT has been the ability of the tool to plagiarize by utilizing sources that are not cited, as well as by creating content that users claim as their own.

After more than four decades of working in two highly regulated industries, healthcare, and higher education, I am not naïve about the timeliness (or lack thereof) of government policymaking and deliberation regarding new technologies.

In July 2024, the Copyright Office issued a report, Copyright and Artificial Intelligence – Part 1: Digital Replicas. The report is the first of several to be issued by the Office in response to its August 2023 Notice of Inquiry (NOI).

The Copyright Office issued a second report, Copyright and Artificial Intelligence—Part 2: Copyrightability, in January 2025. The next report (Part 3) will consider the implications of training AI models on copyrighted works, licensing considerations, and the allocation of any liability.

I believe that each of these reports is important to understand for any person or entity utilizing AI tools for their job, for personal or institutional research, or for any other purpose. I chose ChatGPT-4o to analyze the pdfs for each report. I posted my prompts as well as the responses in the separate sections that follow.

My first prompt and the AI’s response are listed below:

WEB:

I have two recently published documents from the United States Copyright Office regarding Artificial Intelligence. Can you provide me with a detailed review and analysis of each one including a prediction as to how these guidelines could potentially impact companies’ future usage of Artificial Intelligence?

ChatGPT said:

Please upload the two documents from the U.S. Copyright Office regarding Artificial Intelligence, and I will provide a detailed review and analysis of each, including potential impacts on companies’ future AI usage.

WEB note: I uploaded the pdf that I downloaded from the Copyright Office. The dialogue that follows is from ChatGPT-4o until I indicate otherwise.

Analysis and Review of the U.S. Copyright Office Report: “Copyright and Artificial Intelligence, Part 1 – Digital Replicas” (July 2024)

The first part of the U.S. Copyright Office’s AI study focuses on the legal and policy challenges posed by digital replicas—realistic AI-generated imitations of individuals’ voices or appearances. This report explores current laws, potential gaps, and the need for federal legislation to regulate the use of digital replicas. Below is a detailed analysis of its key points, followed by an assessment of how these recommendations could impact businesses utilizing AI.

Key Takeaways from the Report

1. The Rise of AI-Generated Digital Replicas

• AI can now clone voices, generate images, and create deepfake videos with unprecedented accuracy and speed.
• High-profile incidents (e.g., AI-generated songs mimicking Drake and The Weeknd, deepfake political robocalls, and unauthorized AI-generated advertisements featuring celebrities) highlight the growing risks of these technologies.

2. Inadequacy of Existing Legal Protections

The report evaluates whether current laws (state and federal) sufficiently address unauthorized digital replicas. Findings include:

State Laws:

• Right of Publicity: Varies widely by state; some require individuals to be celebrities or have a commercially valuable identity.
• Privacy Laws: Only apply to personal harm, and enforcement is inconsistent.
• Recent Legislation: States like New York and Louisiana have begun passing laws regulating digital replicas.

Federal Laws:

• Copyright Law: Protects fixed works, but not an individual’s voice, likeness, or artistic style.
• Lanham Act: Covers false endorsement but requires proving consumer confusion.
• Federal Trade Commission (FTC) Act: Prohibits deceptive practices, but AI-created digital replicas often fall outside its scope.
• Communications Act: Covers AI-generated robocalls but does not broadly regulate digital replicas.

3. Call for Federal Legislation

The Copyright Office concludes that a new federal law is needed to address unauthorized digital replicas, proposing the following key elements:

Scope of Protection

• The law should cover all individuals, not just celebrities.
• It should target realistic (not exaggerated or stylized) digital replicas that closely mimic real people.

Term of Protection

• Protection should last at least the individual’s lifetime.
• Postmortem rights should exist for a limited period if the individual’s persona continues to be commercially exploited.

Licensing & Consent

• Individuals should have the right to license their digital replicas.
• Licenses should be limited in duration (e.g., 5-10 years) and not assignable outright to prevent exploitative agreements.
• Minors should have additional safeguards, such as court approval for licensing agreements.

D. Liability for Unauthorized Use

• Primary Liability: Applies to those who distribute or profit from unauthorized digital replicas.
• Secondary Liability: Platforms or services that facilitate unauthorized digital replicas may be liable, but safe harbors could apply if they respond promptly to takedown requests.

E. First Amendment Considerations

• The law should balance free speech with individuals’ rights.
• Fair use exemptions for news, parody, and artistic expression should be considered, but deceptive or exploitative uses should be restricted.

F. Remedies

• Statutory damages and injunctive relief should be available.
• Criminal penalties could apply for severe cases, such as AI-generated deepfake pornography or fraudulent impersonation.

Potential Business Impacts of These AI Regulations

1. AI Content & Media Companies

• Increased compliance burden: Companies offering voice cloning, deepfake video generation, or AI-based advertising will likely need to verify user consent and track licensing agreements.
• Higher legal risks: Unauthorized use of public figures’ voices or images could lead to lawsuits or fines.
• More expensive licensing: Companies may need to pay for licensing real voices or images rather than relying on AI-generated imitations.

2. AI-Generated Marketing & Advertising

• Brands must avoid deceptive AI-generated endorsements.
• AI-generated influencer marketing may face tighter disclosure and consent requirements.
• Deepfake-based ads could require pre-screening to ensure compliance.

3. Streaming & Music Industry

• Music platforms (e.g., Spotify, YouTube, Apple Music) may need stricter policies on AI-generated songs using cloned voices.
• AI music tools (e.g., Soundraw, Boomy) will need licensing agreements for AI-generated content mimicking real artists.

4. Social Media & AI Content Hosting

• Platforms (e.g., TikTok, Instagram, YouTube, X/Twitter) may need stronger content moderation to remove unauthorized AI-generated deepfakes.
• Safe harbor provisions (similar to the DMCA for copyright) may shield platforms if they respond promptly to takedown notices.

5. AI Startups & Developers

• New market for AI compliance tools: Companies might develop AI verification systems to check for unauthorized digital replicas.
• Shift to opt-in AI training: AI developers may require explicit permissions to use celebrity or individual likenesses in training datasets.

Predicted Future Trends

• Industry Adoption of AI Ethics & Compliance Standards

• Similar to GDPR for data privacy, companies may proactively adopt self-regulation to avoid litigation.
• AI watermarking or metadata tracking may become industry norms to verify authorized AI-generated content.

• More AI Licensing Deals

• Expect AI-cloned celebrity voice & image licensing agreements (e.g., similar to James Earl Jones licensing his AI voice for Darth Vader).
• Entertainment unions (e.g., SAG-AFTRA) will likely negotiate AI contracts for actors and voice artists.

• Federal Law Will Set a Baseline but Allow State-Level Variations

• The federal law may not preempt state laws, meaning companies will need to comply with both federal and state regulations.
• California, New York, and Tennessee may impose stricter protections due to their large entertainment industries.

Final Thoughts

This first report from the U.S. Copyright Office signals a major shift in AI regulation. If federal legislation follows the proposed recommendations, businesses that use AI to create, distribute, or host digital replicas will need to:

• Adopt stronger compliance measures to ensure digital replicas are properly licensed.

• Monitor AI-generated content to avoid lawsuits over unauthorized likenesses.
• Adapt AI business models to new legal and ethical boundaries.

The second part of the report is expected to address copyrightability of AI-generated works, AI training on copyrighted material, and licensing models. Once available, I can provide further analysis.

Would you like me to summarize key points in a more concise format or compare this report with another AI regulation framework?

WEB note: I did not ask ChatGPT-4o to summarize its key points in a more concise format or to compare this report to another. Instead, I uploaded the second report and added this prompt:

Attached is the Part 2 report. Please provide an analysis similar to what you did for Part 1.

What follows is the result of its analysis.

Analysis and Review of the U.S. Copyright Office Report: “Copyright and Artificial Intelligence, Part 2 – Copyrightability” (January 2025)

The second part of the U.S. Copyright Office’s AI study examines the copyrightability of AI-generated content—a critical legal and policy issue. It clarifies the conditions under which AI-assisted works may qualify for copyright protection, reiterates the human authorship requirement, and considers arguments for and against new legal protections for AI-generated material. Below is a detailed analysis of its key points, followed by an assessment of how these findings could impact companies that create, distribute, or use AI-generated content.

Key Takeaways from the Report

1. The Copyrightability of AI-Generated Works

The Copyright Office takes a firm stance:

• Copyright protection requires human authorship. AI-generated works without sufficient human involvement are not copyrightable.
• AI-assisted works can be copyrighted, but only if human creativity is a significant factor in their creation.
• Prompts alone are not enough—even detailed prompts do not provide enough human control over AI-generated outputs to claim copyright.
• Editing and modifying AI-generated content can make a work eligible for copyright protection, but only if the human modifications meet the originality standard.

2. Legal Framework and Human Authorship

The report examines existing U.S. copyright law and concludes that current legal principles are sufficient to address AI-related copyright issues. The analysis focuses on:

• The U.S. Constitution’s Copyright Clause, which protects human authorship only.
• Judicial Precedent, including:
• Burrow-Giles Lithographic Co. v. Sarony (1884), where a photograph was deemed copyrightable because of the human photographer’s creative choices.
• Thaler v. Perlmutter (2023), where a court upheld the Copyright Office’s rejection of a work created entirely by AI.
• Feist Publications v. Rural Telephone Service (1991), which confirmed that originality—not mere effort—is required for copyright protection.

3. When Does AI-Generated Content Qualify for Copyright?

The report categorizes human contributions to AI-generated content into three main types:

Prompts

• Prompts alone are not enough for copyrightability.
• AI-generated outputs remain highly unpredictable and lack direct human control.
• The Office draws an analogy to hiring an artist to execute an idea—without the ability to control the execution of expressive elements, the prompter cannot claim authorship.

Expressive Inputs

• If a human inputs an original work (e.g., an illustration, text, or sound recording) into an AI system, and the output retains identifiable elements of the original, the human retains copyright over those parts.
• Example: If an artist hand-draws a design and uses AI to add color, the drawing remains protected, but the AI-generated colors do not receive copyright protection.

Modifications or Arrangements

• A human selecting, modifying, or arranging AI-generated material in a creative way may claim copyright over their contributions.
• Example: A comic book combining human-written text with AI-generated images could qualify for copyright, but only the human-authored parts would be protected.

4. International Approaches to AI and Copyright

The report surveys how other countries are handling AI-generated works:

• Most countries agree that copyright requires human authorship (e.g., South Korea, Japan, China, the EU, and the U.K.).
• Some jurisdictions, such as the U.K. and China, have statutes or court rulings that extend protection to AI-generated works under specific conditions.
• The U.K.’s Copyright Act (1988) uniquely grants copyright to computer-generated works, but the British government is reassessing this policy.
• The Copyright Office does not recommend following the U.K. model and argues that human authorship should remain the standard in the U.S.

5. Arguments for and Against Legal Change

The report evaluates arguments for and against new legal protections for AI-generated works.

Arguments for Copyrighting AI-Generated Works

• Incentivizing AI innovation: Supporters argue that AI companies and users need copyright protection to encourage investment in AI tools.
• Expanding creative opportunities: Some claim that allowing AI-generated content to be copyrighted would empower creators, especially those with disabilities.

Arguments Against Copyrighting AI-Generated Works

• No incentive needed for AI: AI does not need economic incentives to generate content—it produces outputs automatically.
• AI-generated content could harm human creators: A flood of copyrighted AI-generated works could dominate markets, reducing the viability of human-created content.
• AI tools are already profitable: Many AI companies already benefit from patent, trade secret, and copyright protection for their software and models.

6. Conclusion: No Legislative Changes Needed (For Now)

The Copyright Office concludes that:

• Existing copyright law is sufficient to address AI-related copyright issues.
• No new laws are necessary at this time, but the Office will monitor technological and legal developments.
• Further guidance will be provided through updates to the Copyright Office’s policies.

Potential Business Impacts of These AI Copyright Regulations

1. AI Content Creation Platforms

• AI companies must be transparent: Platforms that generate text, images, music, or videos cannot claim copyright on behalf of their users unless significant human input is involved.
• Users cannot rely on copyright for AI-generated works: AI companies may need new business models that focus on licensing AI content rather than relying on copyright protections.

2. Creative Industries (Music, Art, Literature, Film)

• Artists and writers must ensure AI-assisted works meet originality standards before registering them for copyright.
• Music streaming services may filter out AI-generated content that does not have clear human authorship.
• Filmmakers and game developers using AI-generated assets may need to document their creative modifications to claim copyright.

3. Digital Marketing & AI-Generated Advertising

• AI-generated ad campaigns without human-authored components may not qualify for copyright protection.
• Companies must carefully track AI-assisted content creation to avoid legal disputes over ownership and authorship.

4. Legal & Compliance Considerations for AI Companies

• AI firms may face challenges in protecting AI-generated assets.
• Users and developers need to disclose AI-generated elements when seeking copyright protection.
• The ruling strengthens the position of copyright holders whose works are used in AI training datasets.

Predicted Future Trends

• Increased Copyright Disputes Over AI Works

• • Expect more lawsuits over AI-generated works, particularly when human authors claim ownership over AI-assisted content.

• New Licensing Models for AI-Generated Content

• • Since AI-generated works may not receive copyright, companies may develop new licensing agreements to protect and monetize AI content.

• Expansion of AI Watermarking and Content Attribution

• • AI-generated works may be required to carry labels or watermarks to indicate they lack copyright protection.

• Stronger Industry Standards for AI-Assisted Creativity

  • Creative industries (e.g., film, publishing, music) may establish best practices for using AI tools without undermining human authorship.

Final Thoughts

The U.S. Copyright Office’s firm stance on human authorship reinforces the importance of originality in AI-assisted creativity. Companies using AI for content creation should:

• Ensure AI is used as a tool, not a creator.
• Disclose AI-generated elements when applying for copyright.
• Adapt business models to rely on licensing and attribution rather than copyright exclusivity.

Would you like further insights on how these findings could impact a specific industry or business model?

WEB note: I did not ask ChatGPT-4o to summarize its key points in a more concise format or to compare this report to another.

A Few Thoughts to Consider

Artificial Intelligence is not a new concept. In fact, its conceptual origins stem from the 1950’s. When the availability and launch of LLMs, specifically ChatGPT, provided access to hundreds of millions of U.S. citizens, regulators like the U.S. Copyright Office increased their focus on potential outcomes.

The Copyright Office’s Notice of Inquiry (NOI) may have been later than expected, but with the publication of two reports over the past seven months, it’s catching up with the technology and the evolving issues.

The major recommendation stemming from the ChatGPT-4o analysis in Part 1 is its recommendation for a federal law addressing unauthorized digital replicas. While you can argue that some entities won’t abide by the regulations, most companies that operate legally in the U.S. will.

In Part 2, the Copyright Office takes a hard stance that copyright protection requires human authorship. More specifically, AI-generated works without sufficient human involvement are not copyrightable. Prompts are not enough. Even better, legislative changes are not needed.

Part 2 will have many applications in education. Faculty who formerly operated under the impression that their syllabi were copyrighted should be careful about how courses are created in the future. AI tools will make course generation easier than ever before. However, my advice is that every course creator should save relevant documentation indicating that the creation of the course was the result of their human initiatives and not due to using a prompt.

College students, undergraduates and graduates alike, will need to document their original ideas and activities that indicate the substantial portion of the written and/or visual output resulted from their initiatives.

I look forward to the Copyright’s issuance of Part 3 which will address copyright issues related to training the AI models, a much discussed and debated topic. Meanwhile, I advise anyone using AI tools to consider the recommendations and analysis in Part 2.

Lastly, I read both of the Copyright Office’s reports before I asked ChatGPT-4o to analyze and summarize. I thought the analysis was commendable and usable. As always, I have chosen to present material created by ChatGPT as AI-created content versus attempting to pass it off as my own.