Fair Use After Warhol v. Goldsmith: Legal Impact

The Supreme Court's Warhol v. Goldsmith decision establishes that for fair use to apply, artistic appropriation must be sufficiently transformative, distinguishing between works that genuinely comment on or critique the original versus those that merely serve commercial purposes; this ruling has significant implications for understanding copyright protection in the era of generative AI, where the line between transformative use and infringement remains contested.

AI-generated works may qualify for copyright, but determining ownership is complex. The US Copyright Office has stated prompts alone are insufficient for copyright protection. UK law recognizes copyright in computer-generated works with 50-year duration (not life plus 70), with ownership defaulting to the person making arrangements. Contracts can override default ownership rules. Some jurisdictions require disclosure of AI-generated content. The US requires disclosure for copyright registration. Adobe is piloting metadata labeling for AI-generated content in Photoshop. Copyright is not a monopoly—creators cannot monopolize all art or music. Copyright only becomes actionable when someone copies the work. The Warhol case addressed whether appropriation art was transformational enough for fair use, with recent court decisions narrowing transformational use protections, affecting AI training defenses.

Generative AI companies commit copyright infringement through multiple mechanisms: scraping content for training datasets, using plagiarized sources, utilizing content during training, and creating substitute outputs. The New York Times case exemplifies this, where their content was used freely without consent. The fair use defense centers on market substitution versus amplification. In Cadri v. Meta, Judge Chabria found fair use focusing on transformative use and societal benefit. In BART's v. Anthropic, Judge Alup distinguished between training on pirated versus legitimately acquired materials, treating piracy as a distinct wrong. The Andy Warhol decision shifted analysis to whether the copyright owner expected payment for the specific use. The Google Books case found fair use because it amplified markets by making works searchable worldwide. There is an existing licensing market for generative AI training, with the New York Times having licensed content for data mining purposes. Factor four of fair use considers whether there is a vibrant market for licensing and adverse impact on that market. The Copyright Office warns about market substitution, where AI tools replace human creators entirely. Beyond copyright, generative AI raises multiple legal issues: trademark dilution through false brand attributions, defamation concerns where AI lacks intent for liability, biometric data and right of publicity issues when AI creates celebrity images, and Section 230 defenses for platforms hosting user-generated infringing content. The music industry has filed major litigation against Suno and Udio, with Suno admitting to scraping the entire internet of music without payment.

This section analyzes how fair use doctrine serves as the primary framework for resolving AI training disputes. Major infringement claims focus on whether using copyrighted material to train models constitutes infringement. Fair use analysis centers on transformative nature (Factor One) and market harm (Factor Four), with other factors being less significant in new technology cases. Transformative use arguments are compelling because AI training creates intermediate uses resulting in entirely new works. The Supreme Court's recent Warhol decision demonstrates how market harm analysis can determine outcomes, suggesting this area will likely reach the Supreme Court within 3-5 years.

The core legal question regarding generative AI training is whether copying copyrighted material for training purposes constitutes fair use. Most lawsuits are at the motion to dismiss stage, as fair use is a defense rather than a pleading requirement. Courts have rejected DMCA Section 1202 claims and derivative work theories. The analysis hinges on four fair use factors: purpose/character, nature of work, amount used, and market effect. Transformative use has emerged as key, with Google Books establishing precedent for transformative use in search engine contexts. The idea-expression dichotomy supports fair use, as training involves learning patterns rather than copying protected expression. Style is not protected, enabling emulation of famous artists' styles. Current precedent suggests transformative use applies, but Warhol case uncertainty remains.

There are two ways to profit from copyright: selling (assignment) or licensing. When you sell, you give all rights away; when you license, you only give some rights and retain the rest. A license automatically expires at the end of its stated term. When publishers ask for assignment, always include a reversion clause that says if the work fails to sell or goes out of print, the copyright comes back to you. Every major social media site has clauses in their Terms of Service that say they own the rights to your posted content. The advice is to not put actual artwork on social media sites—instead, post a notice with a link to your website where you control all the copyright. When consigning works to galleries, artists should dictate terms of sale, including a floor price that galleries cannot sell below without permission. Galleries can sell the painting itself but cannot sell the copyright. When working with agents, ensure they cannot bind you to deals without consent and include language allowing either party to terminate on 30 days notice. Creative businesses must pay taxes on any profit made, even if only one dollar. The difference between a hobby and a business comes into play when you're losing money—businesses can deduct expenses like materials, travel, and education. Losses can be carried forward indefinitely. Artists should never sign a work made for hire clause. When you create work as an employee (W-2), the employer owns all rights. However, this doctrine does not apply to independent contractors (1099 basis). If you create work for someone as an independent contractor and don't assign the copyright, you still own it. This was established in Reed v. Reed (1989), where the Supreme Court ruled that paying for a sculpture doesn't buy the copyright. If you take somebody else's copyrighted work and publish it yourself or include it in another work, you are infringing that copyright. In 'The Wind Done Gone' case (2000), an author wrote a book based on 'Gone with the Wind' from the perspective of an African-American slave. The court ruled this was not a derivative work because the author did not copy any of Margaret Mitchell's words or character traits—it was an entirely separate work.

Artists entering licensing agreements must understand five critical elements: (1) Identify the parties involved and how assets are held (personally, through a business, or holding company), as this affects risk exposure; (2) Clarify what is being licensed, including derivative works like prints or products; (3) Define the scope including time limitations (one year, three years, or perpetuity), media types (prints, electronic, digital), and distribution channels; (4) Consider value assessment, which is subjective and often undervalued by artists, including both financial and non-financial benefits like social media exposure; (5) Recognize that licenses require strategic thinking about how the arrangement supports long-term goals and allows for termination if not generating expected results.

Wikia uses CC BY-SA (Attribution-ShareAlike) license, allowing reuse with attribution and same-license requirement. Other variants include CC BY-NC (non-commercial) and CC BY-ND (no derivatives). Best practices include verifying source license compatibility, using content to illustrate rather than dominate articles, checking local community policies, using licensing dropdowns for images, and articulating fair use reasoning. When infringement occurs, consider fair use applicability. Wikia receives DMCA complaints and removes infringing content. Admins should warn users and explain violations, assuming good faith. The worst-case scenario is typically DMCA takedown.

When creating derivative works (such as drawings based on photographs), you must obtain permission from the original creator before publishing, selling, or distributing your work online; using licensed images with commercial and modification rights is the safest approach, and selling derivative works requires additional authorization beyond the original image's license.

Copyright gives the exclusive right to publish, make copies, profit from derivative works, edit or alter the work, and sue infringers. Licensing is better than selling because it maintains ownership. When selling to publishers, authors should include reversion clauses that return copyright if sales fall below a certain level. Social media platforms like Facebook and Instagram have terms that state users assign their copyright to the platform, so artists should post on their own websites and link to social media instead. Business losses can be deducted against all income, while hobby losses can only be deducted against hobby income. Employees who create work as part of their job automatically create 'work made for hire' - the employer owns the copyright. Independent contractors retain copyright unless they formally assign it.

The first fair use factor examines whether the new work is transformative and whether it is commercial or non-profit. Transformative works add something new to an existing work, endowing it with new expression, meaning, or message. Commercial use weighs against fair use, but this factor is heavily influenced by the transformative nature of the work. The more transformative the work, the less significant commercial use becomes.

The first fair use factor examines commercial nature and whether the use adds new expression, meaning, or message. While commercial use weighs against fair use, it is not dispositive. The transformative nature of the use is the most important factor and can potentially overrule other factors. A transformative use creates something new and different from the original, which is crucial for fair use determination.

The first factor in fair use analysis examines whether the use is commercial or non-commercial. Commercial use weighs against fair use, while non-commercial, educational, or transformative use weighs in favor. In this case, the defendants' website was not charging for membership, did not advertise, and was intended to inform the public about weddings. The website also displayed the defendants' personal information (name and phone number) for wedding services. This factor did not strongly favor fair use since the website had commercial elements.

The first factor in fair use analysis examines the purpose and character of the use. Nonprofit educational uses weigh heavily in favor of fair use, while commercial for-profit uses weigh against it. Courts increasingly recognize transformative uses—even commercial ones—as fair if they add new meaning or purpose to the original work. Transformative use is a key consideration that can override the commercial nature of a use in favor of fair use determination.

The first fair use factor examines whether the use is commercial or non-profit educational. Uses that critique, comment on, or poke fun at the original work (such as parody) tend to support fair use findings. News reporting, teaching, scholarship, and research purposes also support fair use. Transformative works—those that add new expression, meaning, or message to the original—are particularly favored under this factor. However, commercial uses generally weigh against fair use. Even non-monetized YouTube videos may be treated as having commercial purposes because they seek views, likes, and subscribers.