The Clumsy Race for Innovation in AI
No one can deny that artificial intelligence (AI) and large-language models (LLM’s) are currently some of the most popular items of discussion. Students, business leaders, teachers, and others are relying heavily on models like OpenAI’s ChatGPT for ideas and assistance with projects. In efforts to provide high-quality content generators and LLM’s to an AI-hungry society, many companies are in the race to develop a product that noticeably outperforms the rest. However, while working to innovate in a new frontier, several of these companies are running into issues with a pre-existing and foundational system of laws in a new and obscure field. This legal system is copyright.
What Is Copyright?
To protect content creators, copyright laws control how their works are produced or used. These laws are important, because they incentivize creators to invest time and money into producing high-quality content by giving them the assurance that their hard work won’t be pilfered by another. Copyright laws apply to made creations, including sound recordings, visual art, films, music, literary work, and more. A work can receive a copyright once it is fixed in a tangible medium. Copyright laws can be avoided if fair use can be applied. Fair use allows limited use of copyrighted material without permission, based on four factors involving the purpose and nature of the content. While courts make final determinations, individuals can assess these factors to make informed decisions and avoid legal risks.
AI Training and Copyright Challenges
As many users and developers know, high-quality AI outputs rely on high-quality content used in training. As the saying goes, trash in, trash out. However, copyright laws make finding large amounts of accessible high-quality content difficult. In a recent casei, Anthropic paid $1.5 billion to settle claims over unauthorized use of books in AI training. Courts distinguished between scanned purchased books (potentially fair use) and pirated copies (not fair use). In Kadri v. Metaii, the judge ruled Meta’s use of pirated books for AI training was fair, noting the plaintiffs didn’t argue market harm. However, their act of pirating works was not protected by fair use. In contrast, Authors v. Anthropic led to a $1.5 billion settlement, partly due to use of pirated content. In lawsuits againstiii Midjourney and Chinese firm MiniMax, studios like Disney and Warner Brothers used infringing outputs as evidence of unauthorized training inputs. Figure 1 shows many AI and copyright-related lawsuits in progress across the world, as of September 12, 2025.
Figure 1. chatgptiseatingtheworld.com
The large number of lawsuits of this nature shows that there is obviously a restriction on the innovation of high-quality AI and LLM’s. The remedy to this restriction will need to come from either adjustments to the copyright laws or a better supply of license-free content that developers can use to train AI.
Infringing Outputs and Legal Risk
AI and LLM users can also be in danger of infringing copyright laws by claiming outputs to be their own work. Some models like ChatGPT try to prevent this issue by refusing to respond to prompts that produce outputs that might violate copyright law.
Are AI Outputs Copyrightable?
Suppose I use a well-thought prompt to have Microsoft Copilot create a beautiful landscape photo for me. Can I register that photo at the copyright office? Or can I at least obtain a copyright on the prompt? The answeriv is no. Below are some recent examples regarding this issue.
- Thaler v. Perlmutter v: AI-generated art was denied copyright.
- Zarya of the Dawn vi: Comic book text and arrangement were registered, but AI-generated images were excluded.
- Colorado State Fair Winner vii: Registration denied due to refusal to disclaim AI content.
- Invoke’s “American Cheese” viii: Successfully registered based on documented human arrangement of AI elements.
Looking to the Future
As evidenced by the many AI and Copyright-related issues, innovation will typically grow and change faster than legal systems can change. This delay can cause the ‘growing pains’ of innovation to grow faster than the legal systems. However, these growing pains can cause positive change. Adjustments to copyright laws for AI-related issues can be explored, allowing for more unrestricted innovation. Alternatively, advancements can be made in producing high-quality non-licensed content for developers to use while training their models. Due to the rapid development and usage of AI models, serious effort will need to be applied to the copyright issues to allow for continued innovation. This task can appear daunting. However, as evidenced by history, society grows as it seeks to solve the problems that arise from innovation
Notes
i NPR, “Anthropic Settlement, Authors & Copyright in AI,” September 5, 2025, https://www.npr.org/2025/09/05/nx-s1-5529404/anthropic-settlement-authors-copyright-ai
ii Arthur Gollwitzer, “Federal Courts Find Fair Use in AI Training: Key Takeaways from Kadrey v. Meta and Bartz v. Anthropic,” JW News & Insights, July 11, 2025, https://www.jw.com/news/insights-kadrey-meta-bartz-anthropic-ai-copyright/.
iii Reuters, “Disney, Universal, Warner Bros Discovery Sue China’s MiniMax for Copyright Infringement,” September 16, 2025, https://www.reuters.com/legal/litigation/disney-universal-warner-bros-discovery-sue-chinas-minimax-copyright-infringement-2025-09-16/
iv Christopher T. Zirpoli, “Generative Artificial Intelligence and Copyright Law,” CRS Legal Sidebar LSB10922 (Washington, DC: Congressional Research Service, July 18, 2025), https://www.congress.gov/crs-product/LSB10922.
v Jayashree Mitra, “No Copyright Protection for AI-Assisted Creations: Thaler v. Perlmutter,” Carlton Fields Insights & Publications, March 26 2025, https://www.carltonfields.com/insights/publications/2025/no-copyright-protection-for-ai-assisted-creations-thaler-v-perlmutter
vi Tony Analla, “Zarya of the Dawn: How AI Is Changing the Landscape of Copyright Protection,” Harvard Journal of Law & Technology Digest, March 6, 2023, https://jolt.law.harvard.edu/digest/zarya-of-the-dawn-how-ai-is-changing-the-landscape-of-copyright-protection
vii “Jason Allen’s AI Art Won the Colorado Fair — But the Feds Say It Can’t Get Copyright Protection,” Colorado Public Radio, September 6 2023, https://www.cpr.org/2023/09/06/jason-allens-ai-art-won-colorado-fair-feds-deny-copyright-protection/
viii Adam Schrader, “How This A.I. Image Became the First to Snag Copyright Protection,” ArtNet News, February 12, 2025, https://news.artnet.com/art-world/invoke-snags-first-ai-image-copyright-2608219