生成式AI重塑非顯而易見性與技術揭露標準

Generative AI Reshaping the Standards of Non-Obviousness and Technical Disclosure

GenAI造成專利法上的系統性衝擊。GenAI強大的跨領域整合能力，將專利法上虛擬的PHOSITA，實質上提升為全能人機混合體，非顯而易見性門檻更難跨越。有肯認PHOSITA概念固有的內建彈性，專利相關法規動態適應新技術毋庸制訂新法；美國專利商標局則提出重大貢獻標準，但可能發生不易操作的問題。GenAI能被用以干擾他人專利申請，但法律障礙在於這些內容缺乏與人類知識的實質連結不符公開取得性，或普遍缺乏實驗數據佐證難以滿足可實施性，由美國最高法院Amgen案及日本特許廳專利審查實務中可得到印證。GenAI的發展可能惡化預言式揭露的濫用，加劇專利叢林現象，對市場競爭有負面影響。為維護專利的品質，專利主管機關應對AI輔助發明採取更嚴格的審查標準。

Generative AI (GenAI) brings systemic impacts to patent law. With its powerful cross-disciplinary integration capacity, GenAI effectively elevates the hypothetical PHOSITA (Person Having Ordinary Skill in the Art) into an omnipotent human-machine hybrid, thereby making the non-obviousness threshold harder to overcome. While the built-in elasticity embedded in the PHOSITA concept allows patent law to dynamically adapt to new technologies without the need for entirely new statutes, the significant contribution standard proposed by the USPTO may encounter enforcement difficulties. GenAI can also be used to interfere with others’patent applications, yet legal barriers arise because its outputs often lack substantive linkage to human knowledge and thus fail to meet the requirement of public accessibility, or because such outputs generally lack supporting experimental data and struggle to satisfy enablement. This is evidenced by the U.S. Supreme Court’s Amgen decision and the JPO’s patent prosecution practices. Moreover, the development of GenAI may worsen the abuse of prophetic disclosure, intensify patent thickets, and negatively affect market competition. To preserve patent quality, patent authorities should adopt stricter examination standards for AI-assisted inventions.