Generative AI Is fuelling a copyright crisis. Can the law keep up?

The 1984 United States Supreme Court ruling in Sony Corp v Universal City Studios was a victory for innovation, but left lingering questions about how copyright law should adapt to future technologies.

In the mid-1970s, Sony Corp introduced the Betamax video recorder, which allowed users to record TV programmes and watch them later. While revolutionary, the technology sparked controversy. Media companies like Universal Studios sued Sony, claiming that recording TV programmes, even for personal use, infringed on copyright.

The 1984 United States Supreme Court ruling in Sony Corp v Universal City Studios ultimately sided with Sony. The court decided the ‘time-shifting’ feature had substantial non-infringing uses, protecting it under the ‘fair use’ doctrine. This landmark decision was a victory for innovation, but left lingering questions about how copyright law should adapt to future technologies.

Today, a similar battle is unfolding with the rise of Generative Artificial Intelligence (Gen AI) models like OpenAI’s ChatGPT, Google’s Bard, Microsoft’s Azure OpenAI, and Meta’s LLaMA, and others. These systems train on vast datasets scraped from billions of pages of text, images, and other content — often without permission. Courts worldwide are now grappling with a critical question: does the ‘fair use’ principle still hold in the AI era, or must it evolve?

Copyright lawsuits in the AI era

From the US to the United Kingdom to India, legal challenges are testing the limits of fair use — a principle originally designed to allow limited use of copyrighted material for education, research, and criticism.

NY Times v OpenAI (US)

Last year, the New York Times accused OpenAI and Microsoft of using its copyrighted material without permission or compensation to train AI models that directly compete with its content. The NYT argued that OpenAI’s LLMs cannot be ‘transformative’, as they generate outputs that mimic and compete with the original articles, often pulling material from behind paywalls. Among the several examples cited by the NYT is the unauthorised use of the publication’s 18-month long Pulitzer Prize-winning investigation into predatory lending in New York’s taxi industry. The NYT asserts that such unauthorised use of its copyrighted content causes immense economic harm by diverting audiences from its original content.

Getty Images v Stability AI (UK)

Early in 2023, Getty Images sued Stability AI, alleging that millions of its licensed images were unlawfully scraped to train its Stable Diffusion model. The company claimed this not only infringed its copyright but also allowed users to generate similar works, committing ‘secondary infringement’. The case could set a global precedent for regulating AI models trained on unlicensed content.

ANI v OpenAI (India)

In November, news agency ANI took OpenAI to court, citing concerns over ‘AI hallucinations’ where fabricated information is falsely attributed to legitimate sources. ANI argued that these inaccuracies not only harmed its reputation but also highlighted the risks of AI systems training on copyrighted material without any oversight.

Authors and artists v AI

In July 2023, individual creators, like comedian and author Sarah Silverman, filed lawsuits against OpenAI and Meta, alleging that their books were used without consent to train language models. These cases highlight broader concerns about how AI systems profit from creators’ work without compensating them.

The evolution of fair use: Transformation or Exploitation?

Fair use, the foundation of copyright law, relies on four factors: the purpose of use, the nature of the work, the amount used, and the impact on the market. Courts have long interpreted this doctrine to adapt to new technologies. In Google v Oracle America, for example, the US Supreme Court ruled that Google’s use of Java APIs was ‘transformative’ because it served a new purpose in building Android. Similarly, in the Google Books case, digitising books to create a searchable database was deemed fair use.

Applying this principle to AI is far more complex. AI companies argue their models generate new outputs rather than copying original works, making them ‘transformative’. However, critics contend that these systems often compete directly with original content, causing economic harm. The NYT maintains that GenAI outputs closely mimic inputs and should not be considered transformative.

Towards a new copyright framework

The stakes in these legal battles extend far beyond individual cases. Like the Betamax debate, GenAI raises questions about how to balance innovation with intellectual property rights. Current copyright laws often fall short when applied to global technologies like AI, leading to a patchwork of national regulations.

Some countries have begun addressing the issue. The European Union’s Data Governance Act proposes creating data marketplaces where high-quality datasets can be legally licensed. Canada’s Artificial Intelligence and Data Act mandates that AI systems be trained on lawful and ethically sourced data with full transparency. In the US, privacy laws like California’s Privacy Rights Act give individuals control over how their data is used, which indirectly impacts AI training datasets.

In China, the cyberspace administration requires companies to report their data sources and ensure they comply with intellectual property and privacy laws. Meanwhile, India’s Digital Personal Data Protection Act, 2023, will require companies to obtain explicit consent before processing personal data, though enforcement is pending until rules are finalised.

A need for international co-operation

While national efforts are commendable, territorial laws are inadequate for global technologies. Cases like Getty v Stability AI demonstrate the need for a harmonised international framework. Organisations like the World Intellectual Property Organization (WIPO) and Unesco could play pivotal roles in establishing guidelines that balance innovation with fairness to creators.

Striking the right balance

The copyright challenges posed by GenAI are more than just legal questions — they reflect societal values of fairness, innovation, and intellectual property. The Betamax case taught us that technological progress and creators’ rights can co-exist with appropriate safeguards. The same lesson applies today.

As Brian Christian, author of The Alignment Problem, aptly puts it: “The question is not just whether we can build systems that are aligned with human values, but whose values we choose to align them with — and who gets to decide.” In the age of AI, transparency and fairness must guide the evolving relationship between creativity and technology.

(Abhishek Patni is a New Delhi-based senior journalist. X: @Abhishek_Patni)

Disclaimer: The views expressed above are the author's own. They do not necessarily reflect the views.