As companies expand their use of artificial intelligence, copyright law has become a central area of risk and uncertainty. Building on a recent Tech Marathon webinar presented by Morgan Lewis, we highlight key takeaways from recent court decisions testing fair use in AI training as well as legislative proposals on copyright disclosure and digital replicas. The landscape is shifting quickly, and technology lawyers should be prepared to help clients evaluate AI-related rights and manage risks in transactions.
Copyright Protection for AI Outputs
US copyright law requires human authorship. Courts and the US Copyright Office have repeatedly held that works created solely by AI systems without meaningful human involvement are not protectable. For example, the DC Circuit affirmed that Stephen Thaler’s “Creativity Machine” could not be listed as an author, and the Copyright Office has excluded AI-generated portions of works such as Zarya of the Dawn from protection.
The line between sufficient and insufficient human involvement remains blurry. Prompting alone is not enough, but cases suggest that human selection, editing, or arrangement that meaningfully shapes the output may support copyright protection.
Infringement Risks and Litigation
Recent lawsuits highlight the copyright risks associated with both training AI and using its outputs. There are currently around 30 major cases pending in federal court concerning AI and intellectual property.
Decisions have shown mixed outcomes to date, but some common threads have emerged:
- Most claims that survive a motion to dismiss have been against direct infringement claims based on unauthorized use for training AI
- Courts are rejecting broad arguments that every output and/or the large language model (LLM) itself constitutes derivative works
- Fair use is a key defense for defendants in training AI cases, and the outcomes are likely to be highly factually dependent
As regards fair use, the courts tend to focus on the following four parameters:
- The purpose and character of the use (is the result (LLM or output) transformative compared to the original work?)
- The nature of the copyrighted work (is plaintiff’s work factual or creative?)
- The amount and substantiality of the use (did defendant’s work copy the heart of the original work or a nonessential portion?)
- The effect of the use upon the market (does the use harm plaintiff’s ability to profit from their work? Is it a market substitution for the original works?)
Legislative and Regulatory Activity
Legislators are actively weighing in: in 2024, the US Congress considered more than 100 federal AI-related bills while states have introduced over 600. Proposals range from the Generative AI Copyright Disclosure Act to the NO FAKES Act to California’s AI Transparency Act. Additionally, the White House’s 2025 AI Action Plan emphasizes innovation, infrastructure, and international engagement, while leaving open questions about copyright and content regulation.
The law on copyright and AI is still developing, with courts and policymakers testing the limits of authorship, infringement, and fair use. Companies should expect continued uncertainty and rapid change in this space.