Kilpatrick’s Charles Gray, who focuses his practice on patent counseling and prosecution of both U.S. and international patent applications, recently joined other firm thought leaders to discuss “From Copyright to Patents: Global IP and Legal Issues in GenAI Innovations” at the 21st annual KTIPS (Kilpatrick Townsend Intellectual Property Seminar).
Speakers examined a broad array of topics, beginning with the latest rulings and Copyright Office guidance that are rapidly reshaping copyright’s fair-use doctrine, and the evolving recognition of AI as an inventor, followed by subject matter eligibility for AI-driven inventions and best practices for enablement and disclosure. The session provided a practical overview of how major patent offices in the U.S., China, Japan, and Europe are addressing the most pressing legal issues at the intersection of GenAI and IP.
Charles provides these key takeaways from the discussion:
1. Challenges for In-House Counsel and Applicants
AI-related inventions present significant challenges for both in-house counsel and applicants. The application of patentability standards for AI is highly variable between different nations and even among U.S. art units, requiring applicants to understand these differences to form a proper patent strategy. In-house counsel face difficulties in parsing invention submissions that include an AI component, especially when resources are limited. Many submissions simply “tack on” AI, which, like adding a computer to a software invention, is insufficient for patent eligibility. To handle AI-related inventions effectively, in-house counsel should train engineers and maintain open communication with them.
2. AI Patent Tools and Their Concerns
The patent space utilizes AI tools for prior art searches, drafting, and prosecution. Examples of these tools include PatSnap, IPRally, and Amplified for prior art searches, and IP Author, Qthena, and DeepIP for drafting and prosecution. Despite their utility, these tools come with significant concerns, such as hallucinations, privacy and confidentiality issues, and the commingling of materials from different entities. Other risks include a lack of quality, improper use, and over-reliance on AI, which can lead to “automation bias”. The guidance suggests restricting “direct content use” and emphasizing that professionals should “own the work product” to mitigate these issues.
3. Patent Drafting and Prosecution Strategies
Effective strategies for drafting and prosecuting AI patents are crucial for success. In drafting claims, it is helpful to recite a neural network (NN) or predictive model, but it is essential to detail its specific use and how it is trained. The specification must describe a practical application and the technical improvements the invention provides. Adding dependent claims that utilize the NN’s output in a broader system is also a key strategy. For prosecution, it is important to tell the invention’s “Story” to the examiner, focusing on how the invention improves a computer or other technology. Resolving rejections with the examiner through interviews is highly recommended over appeals.
