.png)
The rapid advancement of generative artificial intelligence (AI) presents a series of opportunities for in-house legal teams, and of complex legal considerations, particularly in the realm of intellectual property (IP) and copyright law. To evaluate the potential rewards and risks of generative AI, LegalOn hosted a webinar on ‘Generative AI For In-House Counsel: Navigating AI Risks in the Business and with Third Parties.’
You can watch a recording of the discussion here.
Speakers included Stanford Law Professor Mark Lemley (and the world’s most-cited scholar on IP issues), Noah Waisberg, the founder of legal tech startup Zuva and the author of ‘AI for Lawyers’, and LegalOn CEO Daniel Lewis. Discussion topics included generative AI in legal work, the adoption of AI in legal teams, real world applications for generative AI, and what became the most debated subject: IP and contractual considerations.
Generative AI operates by learning from vast datasets, often encompassing a range of copyrighted materials. This training process, crucial for the AI to gain understanding and generate novel outputs, raises significant questions in IP law. As highlighted by Professor Lemley, the question isn't just about the AI’s output but extends to the fundamental process of how these AI systems are trained. Are we, in training AI with copyrighted materials, inadvertently walking a tightrope over the chasm of copyright infringement?
Given the events that led to the evolution and creation of generative AI, there has been a flurry of lawsuits from content creators claiming that their copyrighted works have been used in AI training datasets without permission. These claims are at the forefront of a growing number of legal disputes, highlighting the need for clarity in how we approach copyright law.
As Lemley highlighted, "We now have close to a dozen lawsuits that say some form of the same thing, which is that you [GenAI models] trained on one of my works."
One prominent example is the case of Getty Images v. Stability AI, where Getty Images alleged that around 12 million of its copyrighted images were used without permission to train Stability AI's image generator. With damages of $150,000 per infringed work, the implications are massive even if only a fraction of claims succeed. As Lemley explained, "Getty Images has claimed 2 million infringements of registered copyright works, among other things. And so there’s differences in damages if they’re registered versus not registered. But if they’re registered, Getty Images is asking for, say $150,000 per violation times 2 million violations, a sum which is existential [for Stability AI’s survival]."
The concept of ‘Fair Use’ is central to the debate on AI and IP. The ‘Fair Use’ legal doctrine allows limited use of copyrighted material without permission for specific purposes like education, research, or commentary. Applying this to AI training raises questions: Is such use 'transformative' enough to qualify as fair use? These gray areas in the law create a landscape ripe for legal interpretation.
According to Lemley, "I think the key question on the training is whether it’s going to be fair use to make these copies for internal use and then not share them with the world." He argues this training use case is analogous to Google's book scanning project that was deemed fair use despite copying millions of books (in the Authors Guild v. Google case).
However as Waisberg countered, "I think when I read the Google case, what I saw was the court really centering on the idea that Google Books was good for authors. That by exposing small snippets of their books and their works to the world, they were kind of bringing attention to these works like Robby the Robot that might not otherwise get noticed." Whether courts see similar transformative value in AI training remains to be seen.
It’s becoming increasingly evident that our current copyright framework was not crafted with AI in mind, leading to challenges in addressing AI's unique nature. There's a growing call for regulation that can accommodate the nature of AI while protecting creators' rights, potentially leading to new legal frameworks designed specifically for AI. In October, the White House issued an Executive Order meant to establish guidelines and standards for the governance of AI.
Legal professionals, particularly in-house teams, must therefore stay informed about the developments in AI and IP law. They play a crucial role in guiding organizations through these complex legal territories, ensuring compliance and proactive risk management.
As LegalOn CEO Daniel Lewis emphasized, "In-house counsel should proactively review provider terms, limit data sharing, and maintain human oversight of AI." Vigilance from legal teams is essential amidst the countless ambiguities.
The nexus of AI and IP law is fertile ground for discussions across technological, legal, and ethical dimensions. As we progress further in the era of generative AI, collaboration among technologists, legal professionals, policymakers, and creators is essential. Together, they can forge pathways that respect creators' rights while unleashing AI's transformative potential. In this journey, innovation, fairness, and respect for IP law's foundational principles should be the guiding lights.
As AI evolves, LegalOn offers a practical solution: efficient, AI-powered contract review software that incorporates practical guidance content drafted by our team of expert attorneys (not robots). Our tool simplifies complex legal challenges, helping in-house teams manage contract review with clarity and speed. Trust LegalOn to help your team stay compliant and efficient in a world where technology constantly reshapes the legal boundaries. Book a meeting with a member of our team to learn more here.
See how LegalOn can save you time, reduce legal risk, and free you from tedious work.
Sign up to request free early access to LegalOn
