Intellectual Property and Generative AI

Many companies utilize AI tools to streamline their content creation processes. While these tools offer substantial support for daily operations, they also pose potential legal risks related to Intellectual Property (IP). To address these challenges, the World Intellectual Property Organization (WIPO) has released a factsheet designed to guide organizations in navigating these complexities effectively.

Generative AI relies on machine learning models. It can produce new content in seconds by constantly processing large amounts of data. This includes text, images, audio, and videos based on specific user prompts. The output frequently results in intricate IP issues, including the revelation of confidential information, breaches of intellectual property rights, and challenges regarding IP ownership. 

Although eliminating these IP risks is not feasible, the considerations outlined below may be helpful for companies and organizations addressing IP matters in this rapidly changing technological landscape.

Training data and confidential information

Gathering and utilizing training data presents one of many potential intellectual property challenges. AI can source its information from public websites or retain and use users’ prompts for training. When companies or organizations develop generative AI tools from the ground up or adjust existing models with confidential data (i.e., trade secrets), there is a potential risk of this information becoming publicly available. Third parties may use prompt injection techniques to retrieve such data, including sensitive company information.

IP rights and ownership

Intellectual property laws in many countries were not created with generative AI in mind, leading to ambiguity regarding protecting AI-generated outputs and owning such rights. While this concern is minimal for certain types of intellectual property, such as trademarks, it poses substantial challenges for copyright.

The U.S. Copyright Office has indicated that copyright protection requires human creativity. Providing user-generated text prompts is not a creative contribution, as these only influence the output produced by the AI. However, in certain instances, users may retain copyright over an output produced by AI if they have adjusted the prompts and settings required to create that work. These differing interpretations lead to legal ambiguity regarding the global acknowledgment of copyright for works created by AI.

IP Infringement

Global litigation is underway to determine if training AI with IP-protected materials and using the resulting models constitutes IP infringement. While these cases mainly focus on copyright and trademarks, other types of IP, like industrial designs, database rights, and patents, could also be involved.

Liability may extend to developers and users, as individuals can be liable for copyright violations regardless of intent. While some countries have IP law exceptions for generative AI, such as fair use and text mining, the lack of international consistency and the ambiguous application create uncertainty.

As generative AI technology continues to evolve, organizations must carefully navigate the complex landscape of intellectual property rights. By understanding the potential risks associated with training data, ownership issues, and the implications of IP infringement, companies can better prepare themselves to manage these challenges. Staying informed and proactive will be crucial in leveraging AI tools while mitigating legal risks.

The information in this article is for general informational purposes only and should not be construed as legal or tax advice. It is not a substitute for consulting with a qualified legal or tax professional.