The use of artificial intelligence (AI) in inventive activity is creating complex issues for patent law, especially around disclosure, inventive step, and inventorship. At the International Association for the Protection of Intellectual Property (AIPPI) Congress held at PACIFICO Yokohama, a session titled “Artificial Creativity or Knowledge Reality: The Use of Generative AI in Pharma and Biotechnology and Its Impact on the Patent Application Process” examined these emerging challenges, emphasizing the urgent need to clarify rules as AI becomes more deeply embedded in the inventive process.

    The increasingly important role of AI in innovation

    The discussion opened with a striking study on AI-generated molecules, in which an AI system was tasked with creating new compounds. In Phase I clinical trials, AI-designed molecules achieved a success rate of 80%–90%, significantly higher than the historical industry average. In Phase II, the success rate was about 40%, roughly in line with historical averages, though the sample size was limited. While AI has excelled at improving drug safety and identifying “drug-like” properties, there remains room for improvement in boosting efficacy.

    Given this, how much information about AI involvement needs to be included in a patent application?

    Views differ on this question, reflecting a lack of global consensus. One view holds that the degree of disclosure should depend on how AI was used. If AI served merely as a tool for tasks such as screening compounds, detailed disclosure may not be necessary. However, if the AI model itself is part of the claims, or if the invention is a new algorithm, then full disclosure is required. This may include:

    The architecture of the AI model;

    The specific training data used;

    The date on which training was completed;

    Use of the AI’s output as supporting evidence for the validity of the invention.

    From Japan’s perspective, current patent law does not set out specific rules for AI-related inventions. The core requirement continues to follow traditional standards—the specification must clearly describe the invention so that a person skilled in the art can carry it out. This pragmatic approach aims to satisfy requirements in most jurisdictions, even in the absence of explicit AI guidance.

    In the United States, the situation is different. In April 2021, the U.S. Patent and Trademark Office (USPTO) issued relevant guidance emphasizing that the duty to disclose material information is a core principle of patent practice. This places responsibility on patent practitioners to be transparent about AI’s role in the invention, as they bear ultimate responsibility for the accuracy of the application materials.

    A major issue raised at the session was reproducibility. The primary concern is that data generated by advanced AI may not be replicable by patent examiners. If an examiner cannot reproduce the process or verify the results, it becomes difficult to assess patentability requirements such as inventive step and sufficiency of disclosure. This challenge goes to the heart of the patent system: a limited monopoly in exchange for full public disclosure of the invention.

    Strategic considerations for practitioners

    There is no one-size-fits-all answer to when to disclose AI involvement; the decision must be made case by case.

    However, a key principle practitioners should follow is ensuring the accuracy of submissions. Maintaining credibility with patent offices is essential. A practical suggestion is to frame discussions of AI’s future potential in the future tense to avoid making unverified assertions.

    Ultimately, for patent attorneys and agents, the most important guiding principle is to be proactive. Simply asking inventors what they invented is no longer sufficient. Practitioners must now probe how inventors used AI during the inventive process. Understanding how AI was used is the first—and most critical—step to addressing these new, complex disclosure requirements.

    Source: China IP Protection Network.