方法專利分離式侵權及其對人工智慧專利之影響

Divided Infringement and Its Impact on Artificial Intelligence Patents

美國聯邦最高法院在著名的Akamai案之中否定了聯邦巡迴法院有關專利分離式侵權行為的見解，再次強調即使是分離式侵權行為仍然必須附麗於直接侵權行為，也就是在依照美國專利法第271條（a）或其他法規範而缺乏直接侵權行為人時，行為人即不會構成第271條（b）之間接侵權行為。聯邦巡迴法院後來所做出的Akamai V判決擴張了直接侵權行為的範圍，成為後續此類案件之準則。人工智慧之相關專利目前一般可歸類為軟體及方法專利，其專利適格性也必須依照Alice/Mayo測試法加以判斷。此外，不論在機器學習的訓練上抑或人工智慧專利的實施過程中都會牽涉到為數眾多的第三人，其與分離式侵權行為遂具有很密切的關係。本文在探討晚近的分離式侵權行為案件之後，簡介歐美在人工智慧專利上所面臨的各種問題，以及與分離式侵權行為間之關係與風險。

"In 2014, the U.S. Supreme Court overruled the Federal Circuit’s decision on divided patent infringement in Akamai, and once again emphasized that, in accordance with Article 271 (a) of the U.S. Patent Law, even divided infringement must depend on direct infringement. In the absence of direct infringers, the defendant will not be found liable for indirect infringement of Article 271 (b). The Akamai V judgment made by the Federal Circuit later on expanded the scope of direct infringement and became the criterion for subsequent cases in this regard. Artificial intelligence related patents can generally be considered as software and method patents. Therefore, their patent eligibility must also be determined according to the Alice/Mayo test. In addition, whether in the training of machine learning or the enforcement of artificial intelligence patents, a large number of third parties are involved, which is closely related to divided infringement. After discussing the recent divided infringement cases, this article introduces the relationship and risks between artificial intelligence patents and divided infringements."