Generative artificial intelligence (AI) models are trained on vast amounts of text, images, code, and music, much of it protected by copyright. Whether that training is lawful, and who owns what these models produce, are among the most contested legal questions in technology.
For businesses, generative AI copyright issues are not only a concern for AI developers. Companies that use AI tools to create content, code, or products also face questions about ownership and liability. This article summarizes where the law stands in late 2026 and how to manage the uncertainty. It is a general overview, not legal advice.
Where Courts Stand on AI Training Data
In the United States, the central question is whether training an AI model on copyrighted works qualifies as fair use. In 2025, federal courts in Bartz v. Anthropic and Kadrey v. Meta found that training generative models on books was fair use, although the Bartz court treated the acquisition of pirated copies as a separate issue (AI Lawsuit Tracker). Anthropic later agreed to a $1.5 billion settlement with authors, which received final approval in July 2026 (Troveo).
Other rulings point in a different direction. In late September 2026, the U.S. Court of Appeals for the Third Circuit affirmed that ROSS Intelligence’s use of Westlaw headnotes to train a non-generative legal research tool was not fair use, while leaving questions about generative AI unresolved (Jackson Walker). Major cases, including consolidated lawsuits against OpenAI, are still pending. Their outcomes could reshape how AI developers license training data.
Approaches outside the United States differ. The Court of Justice of the European Union (CJEU) is considering its first generative AI copyright case, which remained pending as of September 2026 (AIMultiple). Other jurisdictions, such as Japan and Singapore, have adopted text and data mining exceptions that are more permissive toward AI training.
Who Owns AI-Generated Content
Ownership of AI outputs raises a separate set of questions. In the United States, the Copyright Office has stated that copyright protects human authorship, so content generated entirely by AI, without meaningful human creative input, is generally not eligible for protection. Works that combine AI output with substantial human selection, arrangement, or editing may be protected in part. The exact line between assisted and AI-generated work is still being defined case by case.
This creates practical risks for businesses. Marketing content, product designs, or software generated mainly by AI may be difficult to protect from copying by competitors. At the same time, an AI output that closely resembles an existing copyrighted work could expose the user to infringement claims, even if the resemblance was unintentional. Code generated by AI tools can also raise licensing questions if it reproduces open-source code with specific license terms.
Managing Generative AI Copyright Risk
Legal uncertainty does not require avoiding generative AI. It does, however, call for deliberate safeguards:
- Review vendor terms: check whether AI providers offer indemnification for copyright claims and what conditions apply
- Prefer licensed data: when training or fine-tuning models, use data the organization owns or has clear rights to use
- Keep humans in the creative process: document meaningful human contributions to content the business wants to protect
- Screen high-visibility outputs: check logos, published content, and code for close similarity to existing works before release
- Set an internal AI use policy: define approved tools, permitted uses, and review requirements
Organizations should also involve legal counsel early, especially for products built on AI-generated content. Rulings expected over the next year may change the risk picture significantly, so policies should be reviewed regularly.
Navigating an Unsettled Legal Landscape
Generative AI copyright law remains unsettled, with courts reaching different conclusions depending on the type of AI system and how training data was obtained. Businesses can still use these tools responsibly by reviewing vendor terms, preferring licensed data, documenting human contributions, and seeking legal advice for high-risk uses. Organizations building AI-powered products benefit from working with an experienced software development team that designs data pipelines and workflows with compliance and traceability in mind.