Data in the Age of AI
May 20, 2025
.png)
At Notable Capital, we believe that staying ahead in the AI era requires a deep understanding of both the technology and also the evolving legal and operational frameworks that govern it. Last week, we hosted a working session with our partners at Fenwick focused on one of today’s most critical topics: Data in the Age of AI.
The conversation brought together executives, legal leaders, and technologists for a candid, practical discussion on how to navigate the complex issues surrounding data, ownership, and risk in an AI-driven world. Drawing on insights from that conversation, we’re sharing key considerations every business leader should keep in mind as AI becomes increasingly central to operations, strategy, and innovation.
When we talk about AI today, we’re not just talking about generative AI tools like ChatGPT or DALL-E. AI today spans a much broader ecosystem, including machine learning, predictive analytics, and countless business applications under the hood. This distinction matters. Companies need to understand what kind of AI they’re using and what data is flowing in and out of those systems before they can make decisions about contracts, compliance, and risk.
Under current U.S. Copyright Office guidance, works created entirely by AI are not eligible for copyright protection. This means that depending on how they are generated, outputs such as AI-generated images, reports, or marketing materials may lack traditional intellectual property protections.
However, important exceptions exist. For example, if a human meaningfully shapes, curates, or arranges AI-generated elements into a cohesive new work, the resulting creation may qualify for limited protection as a compilation under copyright.
The practical takeaway: businesses should assume that AI-generated outputs are unprotected IP unless they can clearly demonstrate significant human contribution or authorship. Flexibility is key, as the legal landscape continues to evolve.
Today, many customers are pushing to own the outputs generated by AI services they pay for and in many cases, providers are agreeing, making this behavior a market norm.
While sharing ownership rights can make sense, it must be done strategically. Companies should carefully carve out their ability to retain and use de-identified, aggregated data outputs to fine-tune their models and improve services.
In other words: granting customers ownership of specific outputs does not necessarily mean forfeiting the broader data learnings that drive product evolution and competitive advantage.
As AI becomes more embedded in critical business functions, customers are demanding more from vendors, including representations, warranties, and even indemnities related to AI system behavior.
Companies must be thoughtful about what risks they can realistically own. Best practice is to limit representations and warranties to elements the company controls directly, and to avoid making broad assurances about third-party foundational models.
Similarly, indemnities should be tightly scoped to prevent exposure to unlimited downstream liability. Strong contractual defenses, including liability caps, carefully drafted exclusions, and precise language, are becoming standard for any business building or deploying AI solutions at scale.
Model distillation and model theft have emerged as major risks in the AI ecosystem. Recent cases, including allegations involving DeepSeek, illustrate how actors can recreate powerful models by systematically querying an existing system and training a new model on its outputs.
Because model parameters and generated outputs may not be protected under traditional copyright law, companies may want to focus on other defenses: trade secret protections, contractual anti-reverse engineering provisions, and, where appropriate, strategic patent filings.
Organizations developing proprietary models, or heavily relying on third-party providers, must proactively address the legal and operational risks of model theft as part of their broader IP strategy.
Understanding how data flows, how ownership works, and how risk shifts is critical to building durable, defensible companies in the AI era. The frameworks are still taking shape, but companies that act early and stay informed are in a position to lead.
