Copyright in AI: key implications from ongoing legal cases

The legal landscape for artificial intelligence (AI) is shifting rapidly, with high-profile copyright cases in the UK and U.S. set to shape the future of the AI value chain. These cases raise fundamental questions for developers, deployers, investors, and data owners.

Below, we provide a snapshot of the implications and what they mean for stakeholders across the AI ecosystem.

Developers: existential questions and uncertain precedent

For foundation AI model developers, the stakes could not be higher. The outcome of these cases may have existential consequences, not only for those building the foundation models but also for the many application developers building on top of them. 

The risk profile for any given AI developer will depend on the specific datasets and training techniques used, as well as a range of other design choices. As a result, a decision in one case will not necessarily set a precedent for others. The legal process will take time to play out, and in the meantime, these developments will add heightened scrutiny on legislative proposals worldwide. 

The U.S., in particular, finds itself at a pivotal moment, with much of the world’s AI training activity and a regulatory landscape seemingly at a crossroads with the Big Beautiful Bill. The bill, among many other things, seeks to impose a ten-year moratorium on U.S. states enforcing AI-related regulations. Please get in touch with any of our AI group to learn more.

Deployers: navigating output risk and indemnities

For those deploying AI models, the picture is more nuanced. The primary concern is the risk of copyright infringement in relation to model outputs—a risk that remains untested, especially after Getty Images dropped its output-based claims against Stability AI. 

Deployers must scrutinize the wording of IP infringement indemnities in their AI agreements. Careful drafting is essential; we have seen many instances where broad indemnities failed to provide the expected protection. In practice, good governance can help manage infringement risk for many use cases, but the legal uncertainty means vigilance is required.

Investors: rethinking valuation and diligence

Investors considering stakes in AI developers—whether at the foundation or application layer—must adapt their approach to valuation and diligence. The evolving legal environment demands a technology- and use case-led analysis of AI risks. 

At A&O Shearman, we have developed a uniquely practical approach to diligencing these risks in M&A transactions, ensuring that our clients are equipped to make informed decisions in a rapidly changing market.

Data owners: the rise of data licensing

Data owners are watching these cases as closely as anyone. The outcome may accelerate the development of a large-scale data licensing market. Over the past year, we have already seen the emergence of a new ecosystem of data licensing deals for AI training, with structures that vary depending on the stage of the AI development lifecycle. This trend is likely to continue as legal clarity emerges.

Beyond copyright: broader legal consequences

It is important not to view copyright in isolation. The practical consequences of these decisions will influence how other areas of law apply to AI. For example, repurposing data for model development raises significant privacy considerations, adding another layer of complexity for all participants in the AI value chain.

Conclusion

The current wave of copyright litigation is only the beginning. As the law evolves, all participants in the AI ecosystem—developers, deployers, investors, and data owners—must stay alert to the shifting risk landscape and adapt their strategies accordingly. 

At A&O Shearman, we continue to monitor these developments closely, providing clients with the insight and practical guidance needed to navigate this complex and fast-moving area.