The U.S. Copyright Office has completed a comprehensive study examining the copyright issues raised by artificial intelligence (AI). This study was to gather information and policy perspectives to understand the current state of copyright law, identify unresolved issues, and assess potential areas for congressional action.
On August 20, 2023, the Copyright Office issued a notice of inquiry, inviting comments on the topic. By the deadline in December 2023, the Office received over 10,000 comments, which are now accessible through regulations.gov. As part of this initiative, the Office is releasing a report in several parts, with Part 1 published on July 31, 2024, focusing on digital replicas. Upcoming parts will address topics such as the copyrightability of AI-generated works, legal implications of AI training on copyrighted materials, licensing considerations, and potential liability.
The study has garnered varied responses from key industry players, including Getty Images, DMLA, Shutterstock, Adobe, Microsoft, and Google, each providing different perspectives on ethical AI, copyright protection, and the balance between innovation and intellectual property rights.
Below is a high-level overview of how certain key points amongst some of the largest DMLA members are aligned, or not aligned. What’s notable is when a party is not included in the group that advocates for a certain position on a given issue.
Aligned Positions
Getty Images, DMLA, Shutterstock, and Adobe:
- Ethical AI and Compensation: These companies stress the importance of using licensed content for AI training and ensuring compensation for creators. Getty Images, Shutterstock, and Adobe highlight their development of ethical AI models that use licensed datasets and provide remuneration for content creators.
- Transparency and Record-Keeping: There is a common call for transparency in the use of training data. Getty Images, DMLA, Shutterstock, and Adobe advocate for maintaining auditable records of the materials used to train AI models and the sources from which these materials are obtained.
- Protecting Copyrights: Getty Images, DMLA, and Shutterstock argue against the use of copyrighted material without authorization. They assert that AI training on unlicensed data constitutes copyright infringement and stress the need for robust copyright protections.
Microsoft and Google:
- Balance and Innovation: Both companies emphasize the potential of AI to advance human knowledge and improve various sectors, advocating for a balanced approach that fosters innovation while respecting intellectual property rights. They acknowledge concerns from the creative community but argue that existing copyright laws can address most issues.
At Odds
-
Fair Use and Training Data:
- Google and Adobe: Advocates for the application of existing fair use doctrines to address AI-related copyright issues. Adobe argues that the use of copyrighted works for training AI models is transformative and should be protected by fair use, drawing analogies to previous fair use cases involving intermediate copying.
- Getty Images and Shutterstock: Oppose the blanket application of fair use to AI training, arguing that systematic and wholesale ingestion of copyrighted material should not be considered fair use. They call for explicit guidelines clarifying that unlicensed use of copyrighted works for AI training is not fair use.
-
Transparency Requirements:
- Google and Microsoft: While recognizing the need for responsible AI development, they are less specific about the mechanisms for transparency and record-keeping compared to Getty Images, Shutterstock, and Adobe, which call for detailed and mandatory disclosure of training datasets.
- Adobe: Supports transparency but also highlights the need to balance it with confidentiality concerns and competitive advantages, suggesting that general information about dataset sources should be disclosed without necessarily revealing specific proprietary details.
-
Legislative Changes:
- Google: Suggests that current copyright doctrines are flexible enough to handle AI issues and prefers judicial interpretation over new legislation.
- Adobe and Getty Images: Advocate for new regulations to ensure transparency and protect copyright holders from unauthorized use of their content in AI training. Adobe, in particular, proposes a new Federal Anti-Impersonation Right (FAIR) to address economic harm from AI tools mimicking an artist’s style.
Specific Points from Each Company
Getty Images:
- Strongly opposes the use of its copyrighted material without permission for AI training.
- Advocates for maintaining detailed records of training data.
- Calls for legislative action to protect against unauthorized use of copyrighted works.
DMLA:
- Supports ethical AI models and the licensing of content for AI training.
- Emphasizes the importance of transparency and maintaining records of training datasets.
- Highlights the risks to the public and creators from unlicensed use of content.
Microsoft:
- Focuses on the potential benefits of AI while acknowledging the need to address copyright concerns.
- Advocates for a balanced approach that encourages innovation and respects IP rights.
- Supports existing copyright doctrines and prefers judicial interpretation over new legislation.
Google:
- Emphasizes the flexibility of existing copyright laws to address AI issues.
- Argues for the fair use of publicly available data for AI training.
- Supports judicial interpretation as the primary means of resolving copyright issues in AI.
Shutterstock:
- Strongly advocates for ethical AI, using licensed data, and compensating creators.
- Calls for explicit guidelines clarifying that the use of unlicensed copyrighted works for AI training is not fair use.
- Supports maintaining detailed records of training data to ensure transparency.
Adobe:
- Supports the use of licensed content for AI training and compensating creators.
- Emphasizes the importance of transparency balanced with confidentiality concerns.
- Argues that the use of copyrighted works for training models is transformative and should be protected by fair use.
- Proposes new regulations, including the FAIR Act, to protect creators from AI tools that intentionally mimic their style for commercial gain.
- Advocates for an international harmonization of copyright protections to support the responsible development and use of AI.
For an in-depth look at the positions of these companies, visit the full docket here.
