AI and Copyright

AI and copyright is the field of law and policy concerned with the interaction between intellectual property rights and generative artificial intelligence. It centres on two questions: whether AI developers may use copyrighted works — books, articles, images, music and films — to train their models, and whether the outputs of those models can themselves attract copyright protection and be owned. Since the emergence of large generative models in the early 2020s, courts, legislatures and rights holders around the world have been establishing answers through litigation, licensing deals and new regulation.^[1][8]

Background

Generative AI models are trained on very large datasets of text, images, audio and video, much of it gathered from the public web and including professionally produced, copyrighted material. Rights holders argue that this training amounts to unauthorised reproduction of their works; AI developers generally argue that training is a transformative use permitted under doctrines such as fair use in the United States, or under text-and-data-mining exceptions elsewhere. The dispute has produced litigation on several continents and, increasingly, negotiated outcomes: content-licensing markets, rights-holder opt-outs and transparency obligations attached to AI providers.^[1][11]

Key Cases and Developments

The highest-profile case is The New York Times Company v. Microsoft and OpenAI, filed in December 2023, which alleges that the companies trained models on millions of Times articles without authorisation and that their products can reproduce portions of copyrighted journalism. OpenAI has argued that its use of the material is fair use and that courts have progressively narrowed the case, while the Times has continued to press claims and in mid-2026 sought to expand its complaint against Microsoft. In September 2026 the United States Department of Justice sided with OpenAI in the dispute, arguing that training on lawfully obtained content is transformative — a position that aligns the US government with the developers' defence.^[1][2][3]

In the United Kingdom, the High Court's November 2025 judgment in Getty Images v Stability AI largely rejected Getty's copyright claims. The court held that the model does not store copies of the training images and is therefore not itself an infringing copy of them, though Getty succeeded on a limited trademark claim over watermark misuse. The judgment, the first major UK decision on generative AI and copyright, was widely read as a partial win for model developers, with appeals and further claims expected.^[4][5]

Music followed a similar arc. In June 2024 the major record labels sued the AI music generators Suno and Udio over training on copyrighted recordings. After more than a year of litigation, Universal Music Group settled with Udio and Warner Music Group settled with both companies in late 2025, announcing licensing agreements and plans to rebuild the platforms around licensed models, with artists able to opt in for use of their voices and compositions. Sony's litigation continued into 2026, leaving parts of the licensing picture open.^[6][7]

Copyright of AI Outputs

A separate question concerns whether AI-generated material qualifies for protection. The United States Copyright Office has maintained that copyright requires human authorship, meaning purely machine-generated content is not registrable, while works produced with meaningful human involvement may be. The European Union's AI Act, adopted in 2024, takes a disclosure-oriented approach: providers of general-purpose AI must publish summaries of the content used to train their models, while questions of ownership remain with national law. Several jurisdictions have converged on the principle that human creative contribution — selection, arrangement, substantial editing — determines whether an output can be protected.^[11][12]

>See Also

• EU AI Act
• AI Ethics
• Generative AI
• AI Watermarking
• PDPA AI Compliance
• Deepfake