For decades, intellectual property (IP) risk was neatly packaged into a single clause in contracts, offering a broad indemnity that promised everything would be fine—until it wasn’t. The rise of artificial intelligence (AI) shattered this illusion, revealing the complexities of IP risks that the old models could no longer contain. By 2025, a significant shift in how contracts are drafted reflected this new reality.

Historically, IP indemnities were based on the assumption of clear authorship, discrete inputs and outputs, and identifiable risks that could be transferred from one party to another. AI, with its continuous learning, evolving models, and probabilistic outputs, has upended these assumptions, making it clear that IP risks in AI are not singular; they are multilayered and complex.

The litigation landscape around AI in recent years did not answer all the questions but highlighted critical areas such as training data, derivative works, and the ownership of outputs. These elements began to feature prominently in legal challenges, pushing the legal field to reevaluate the traditional IP indemnity clause.

As a response, legal contracts have started to adopt a more structured approach. The monolithic IP clause has given way to a more detailed and honest division of rights and obligations, reflecting the actual workings of AI systems. This includes separate stipulations for input rights, explicit training rights, and specifically outlined output rights, along with new obligations for labeling and attribution.

This segmentation of IP clauses does not merely add complexity for complexity's sake—it represents a more transparent and realistic approach to handling IP risks, acknowledging that different stages of the AI lifecycle pose different risks. Modern contracts now negotiate these risks piece by piece, tailoring indemnities to more accurately reflect the nuanced reality of AI technologies.

For legal practitioners, this change means that IP negotiations have become more challenging. Both clients and vendors find themselves navigating a landscape where risks are not only more difficult to define but also harder to promise away with a single clause. The focus has shifted from broad indemnities to specific allocations of risk across various aspects of AI development and deployment.

Looking ahead, there’s no returning to the old ways. The legal sector is moving towards a model where contracts are not just written but architected, with a keen eye on aligning legal stipulations with the technical realities of AI. This approach is not just about legal compliance but about crafting agreements that are robust, clear, and capable of withstanding the challenges posed by advanced technologies.

As we continue into this new era, the evolution of IP clauses in the context of AI is a testament to the dynamic nature of law and its ability to adapt to the rapid advancements in technology. The era of simple, one-size-fits-all solutions is over, ushering in a period of legal innovation that mirrors the sophistication and complexity of the technologies it seeks to regulate.