The US Copyright Office AI Report is a 2024 document issued by the United States Copyright Office (USCO), a part of the Library of Congress, addressing the intersection of copyright law and inteligência artificial. The report examines how existing copyright principles apply to AI-generated works, the use of copyrighted materials in training aprendizado automático models, and the legal treatment of deepfakes. It provides policy recommendations to Congress, reflecting the office's role in advising on copyright matters.
The USCO, headed by the Register of Copyrights, is responsible for administering the national copyright system. The report emerged from a broader inquiry initiated in 2023, following rapid advances in IA generativa technologies. It is part of ongoing efforts to update copyright law in response to technological change, building on the office's historical functions of registration, recordation, and policy advice.
Background and Legal Context
The USCO operates under the Copyright Act of 1976, which requires only fixation of an original work in a tangible medium for federal copyright protection. Registration, while optional for ownership, is a prerequisite for infringement lawsuits. The office examines applications, records documents, and maintains the copyright catalog, which includes over 16 million online entries post-1977 and nearly 45 million physical cards from 1870 to 1977.
The AI report was prepared in the context of the office's statutory duty to advise Congress on copyright policy. It follows a series of public hearings and comment periods that gathered input from stakeholders, including creators, technology companies, and legal experts. The report addresses questions about whether AI-generated outputs can be copyrighted, the scope of fair use for training data, and the need for new protections against unauthorized digital replicas.
Key Findings on AI-Generated Works
The report concludes that existing copyright law can accommodate some AI-assisted creations, but the degree of human authorship is critical. Works generated entirely by AI without meaningful human input may not qualify for copyright protection, as copyright requires human creativity. However, when a human selects, arranges, or modifies AI outputs in a sufficiently creative way, the resulting work may be protectable.
The USCO applied this principle in its registration practices, requiring disclosure of AI involvement in applications. The report recommends that Congress clarify the standard for human authorship in AI contexts, potentially through legislation. It also notes that the redes neurais and aprendizado profundo models underlying AI systems do not themselves hold copyright, as they are not authors.
Training Data and Fair Use
A central issue in the report is the use of copyrighted works to train AI models. The report analyzes whether such use constitutes fair use under Section 107 of the Copyright Act, considering factors like purpose, nature, amount, and market impact. It acknowledges that training often involves copying entire works, which may weigh against fair use, but also recognizes transformative uses in some cases.
The report does not provide a definitive answer, instead recommending a case-by-case analysis. It suggests that Congress consider a licensing framework for training data, potentially through collective licensing or compulsory mechanisms. The report also highlights the role of modelos de linguagem de grande escala developers, such as OpenAI, Anthropic, and Google DeepMind, in shaping industry practices, though it does not single out specific companies.
Deepfakes and Digital Replicas
The report addresses the growing threat of deepfakes - realistic AI-generated audio, video, or images that can impersonate individuals. It notes that existing copyright law offers limited protection for people whose likeness is used without consent, as copyright does not cover personality rights. The report recommends that Congress enact a federal right of publicity to protect against unauthorized digital replicas, complementing state laws.
It also discusses the intersection with the Digital Millennium Copyright Act (DMCA), which provides a notice-and-takedown system for infringing content. The report suggests that the DMCA may need updates to address AI-generated content, including mechanisms for labeling and verification. It emphasizes the need for balance between free expression and privacy, particularly for public figures and private individuals.
Policy Recommendations
The report offers several recommendations for Congress. These include: clarifying the human authorship requirement; establishing a federal right of publicity for digital replicas; promoting voluntary labeling of AI-generated content; and supporting research on technical measures like watermarking. It also recommends that the USCO continue to monitor developments and issue further guidance.
The report avoids recommending broad changes to fair use, instead favoring targeted adjustments. It calls for stakeholder collaboration and international coordination, as copyright issues are global. The report reflects the office's view that copyright law should remain technology-neutral, adapting to new tools without stifling innovation.
Reception and Impact
The report has been influential in policy debates, cited by lawmakers and industry groups. Some creators have praised its recognition of human authorship, while AI companies have expressed concerns about restrictions on training. The report has also informed discussions in other jurisdictions, though it is specific to U.S. law.
As of 2025, the report remains a key reference for ongoing litigation and legislative efforts. The USCO continues to accept public comments and may issue supplementary reports on related topics, such as the copyrightability of AI-generated music or visual art. The report's impact is likely to grow as AI technologies evolve.
See Also
- Inteligência artificial
- IA generativa
- Aprendizado automático
- Deepfake
- Copyright Act of 1976
References
This article is based on the US Copyright Office AI Report and related public records. For further details, consult the official USCO website and the Library of Congress archives.