AI Copyright & Legal Disputes 2025: Global Court Rulings and Creator Protection

AI-Generated Works & Copyright Disputes: Global Legal Issues & Creator Protection

As generative AI tools become more sophisticated, they increasingly generate text, art, music, and more. But who owns those outputs? Can AI creations get copyright? And when creators supply training data, what legal protections do they have? In this post, we examine global legal precedents, statutory frameworks, and practical protection strategies for creators in the AI era.

1. Key Legal Concepts & Challenges

Before diving into cases, let’s outline the fundamental legal tensions in AI copyright:

• Human Authorship Requirement — Many jurisdictions hold that copyright requires human creative input; pure AI creation may not qualify.
• Use of copyrighted works for training — AI models are typically trained on large corpuses, raising issues of reproduction, derivative works, and fair use/data mining exceptions.
• Output infringement — If AI output is substantially similar to a copyrighted work, it could infringe protected expression.
• Fair use / TDM — Some regions allow text and data mining (TDM), but its application to AI remains unsettled.
• Memorization risks — Research shows large AI models may reproduce copyrighted text verbatim, leading to liability concerns.

2. Major Court Decisions & Precedents

2.1 U.S. – ROSS Intelligence v. Thomson Reuters (Ongoing)

The ongoing ROSS Intelligence v. Thomson Reuters case in Delaware questions whether ROSS’s use of Westlaw headnotes for AI training constitutes fair use, with no final ruling as of 2025.

2.2 U.S. – California Fair Use Cases

Several California courts accepted fair use defenses in AI cases, stressing transformation and lack of market harm, though decisions vary.

2.3 U.S. – Thaler v. Perlmutter (Human Authorship)

Courts reaffirmed that works created solely by AI cannot qualify for copyright protection without human input.

2.4 U.S. – Getty Images v. Stability AI (Ongoing)

Getty Images sued Stability AI, alleging unauthorized use of copyrighted images for AI training, highlighting risks for AI image generators.

2.5 UK & EU Approaches

The UK applies “fair dealing” rules, while the EU explores broader copyright reforms under its AI Act. Many global jurisdictions are still developing their frameworks.

3. Creator Protection Strategies

• Offer clear licensing terms for dataset usage.
• Use watermarking & metadata to prove provenance.
• Register works and maintain process documentation.
• Define contractual terms with AI platforms or partners.
• Adopt technological protections (e.g., restricted-use APIs).
• Pursue litigation selectively when clear copying occurs.
• Support policy advocacy for creator remuneration and AI regulation.

4. Emerging Research & Issues

• Studies show LLMs can “memorize” copyrighted works, complicating infringement analysis.
• Scholars debate whether TDM exceptions apply fairly to AI training datasets.
• Legal academics caution against granting “special treatment” to AI under copyright law.

5. Summary & Outlook

The copyright status of AI-generated works remains contested worldwide. Courts agree on one point: human authorship is essential. AI training may or may not qualify as fair use depending on jurisdiction and purpose, and AI outputs resembling protected works may trigger liability.

For creators, proactive licensing, technological protections, and legal documentation are essential. For policymakers, balancing innovation with creator protection remains the core challenge.

References & Further Reading:

• U.S. Copyright Office — Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence (2023)
• Reuters — Thomson Reuters Sues ROSS Intelligence Over Westlaw Data Use (2020)
• Jones Day — AI Works and Human Authorship Guidance (2023)
• Ropes & Gray — Artificial Intelligence and Copyright Law (2024)
• The Guardian — Getty Images Sues Stability AI for Copying 12m Photos (2023)
• arXiv — Copyright Violations and Large Language Models (2023)