The rapid development of artificial intelligence (AI) is presenting unique challenges in the realm of patent applications. As AI is increasingly utilized in various industries, questions are emerging about how AI-related activities are classified under existing patent law frameworks, particularly with regard to prior art.
Two critical questions have come to the forefront: whether disclosing information to AI constitutes a public disclosure, and whether AI-generated information itself can be considered prior art. According to U.S. patent law, prior art entails publicly available knowledge that can disqualify a patent application if it reveals or makes the claimed invention obvious.
Disclosing to AI
In the context of large language models, AI consumes vast amounts of text to facilitate more human-like interactions. Given this, developers are incessantly feeding more data into AI systems to enhance their performance. This has led to questions regarding whether the information used to train AI effectively becomes public disclosure, thereby potentially invalidating future patents.
Prior art includes printed publications, public uses, or any information that is publicly accessible. The accessibility of such information, even if no one directly sees it, renders it prior art if it can be found with reasonable diligence.
A user’s input into an AI system may be viewed as a disclosure since that information is utilized to make the AI “smarter.” This makes the information public in a similar sense to oral presentations that become prior art because they are publicly known.
AI-Generated Information
Beyond the training materials, AI produces significant amounts of content autonomously. Such AI-generated information might be considered prior art as a printed publication, public use, or otherwise publicly available data. The status of AI-generated information as prior art hinges on its public accessibility and reproducibility.
If AI-generated responses are stored in a searchable database, they may qualify as printed publications. However, if these outputs are ephemeral and inaccessible after the initial generation, they likely won’t meet the criteria for prior art. The challenge is further complicated by the accuracy and operability of the AI-generated information.
It’s noteworthy that some AI outputs have been found to be fabricated or inaccurate, raising concerns about their viability as prior art. Until AI platforms demonstrate consistent reliability, patent applicants will need to scrutinize AI-generated information to argue against its operability as prior art.
This evolving intersection of AI and patent law underscores the need for legal professionals to stay vigilant and informed about how AI can impact patent environments. For additional insights, the full discussion is available at Bloomberg Law’s article.