背景
人工智能作為創(chuàng)作、法律研究和盡職調(diào)查的工具，可以降低成本并提高服務(wù)質(zhì)量，但也帶來了輸出不準確和保密性問題等風(fēng)險。
為了應(yīng)對這些問題，美國專利商標局于2024年2月13日發(fā)布了《人工智能輔助發(fā)明的發(fā)明權(quán)指導(dǎo)意見》并于2024年4月11日發(fā)布了《在美國專利商標局前使用人工智能工具的指導(dǎo)意見》，旨在告知知識產(chǎn)權(quán)從業(yè)人員使用人工智能的相關(guān)風(fēng)險，提供減少風(fēng)險的建議，澄清有關(guān)發(fā)明權(quán)的問題，解決保密和國家安全方面的擔(dān)憂，并強調(diào)誠實和善意的義務(wù)。
根據(jù)美國專利法第35 U.S.C. 115和116條，發(fā)明權(quán)僅限于自然人。美國專利法將發(fā)明者定義為人類個體，強調(diào)構(gòu)思的心理過程，這是人工智能無法完成的。然而，人工智能可以輔助發(fā)明。這意味著，雖然人工智能系統(tǒng)不能作為專利的發(fā)明者，但使用人工智能并不會使對發(fā)明做出重大貢獻的自然人失去被列為發(fā)明者的資格。這確保了專利系統(tǒng)激勵人類創(chuàng)造力，并為對人類有重大貢獻的人工智能輔助的發(fā)明提供保護。美國專利法規(guī)要求列出對至少一項專利權(quán)利要求做出貢獻的所有發(fā)明者，特別是那些對發(fā)明構(gòu)思做出貢獻的人。對于人工智能輔助發(fā)明，每個被列名的發(fā)明者必須做出“重大貢獻”，這意味著他們必須顯著地參與發(fā)明的構(gòu)思或?qū)嵺`，將發(fā)明付諸實施，或?qū)λ鲝埖陌l(fā)明做出有意義的貢獻，超越僅僅解釋已知概念的范圍。重大貢獻是通過Pannu因素來確定的，這些因素評估發(fā)明者在構(gòu)思或?qū)嵺`中的作用。這適用于使用人工智能的單個發(fā)明者和聯(lián)合發(fā)明者。如果人工智能系統(tǒng)被列為發(fā)明者，該申請將被駁回。如果僅僅對人工智能系統(tǒng)保持“知識上的支配”而沒有對發(fā)明的概念做出重大貢獻，那么使用該AI系統(tǒng)創(chuàng)造的任何發(fā)明的個人不會成為該專利申請的發(fā)明人。美國專利商標局提供了如下指導(dǎo)，幫助確定自然人對人工智能輔助發(fā)明的貢獻是否“重大”以確定發(fā)明權(quán)。關(guān)鍵原則包括：1. 人工智能的使用：僅僅使用人工智能并不會使某人成為發(fā)明者，除非他們顯著地影響了人工智能的輸出。2. 問題的識別：單獨識別一個問題并不構(gòu)成發(fā)明權(quán)，除非該自然人顯著地引導(dǎo)人工智能找到解決方案。3. 實踐化：僅僅將發(fā)明付諸實施是不夠的，但顯著提升人工智能輸出會符合資格。4. 基本構(gòu)建模塊：創(chuàng)建關(guān)鍵組件或設(shè)計人工智能以解決特定問題可具有重大意義。5. 知識上的支配：對人工智能的控制在沒有實質(zhì)性發(fā)明貢獻的情況下不會被賦予發(fā)明權(quán)。商標從業(yè)者必須避免使用人工智能生成的圖像作為商標使用的證據(jù)，因為這些圖像不顯示商標的實際使用情況，因此被視為事實陳述錯誤。此外，在美國商標審判和上訴委員會的程序中使用人工智能證據(jù)時應(yīng)持謹慎態(tài)度，以避免無意中提交不相關(guān)的信息，從而被視為不當(dāng)而被駁回。計算機工具，包括生成式人工智能，越來越多地被用于起草提交給美國專利商標局（USPTO）的文件，如專利權(quán)利要求和答復(fù)通知書。雖然沒有禁止使用人工智能進行此類工作的規(guī)定，但所有提交的文件必須由提交文件的人進行審查和簽名，以確認其準確性并符合USPTO的規(guī)定，特別是37 CFR 11.18(b)條款。這包括核實人工智能生成內(nèi)容的準確性并確保其符合法律和證據(jù)標準。所有提交給USPTO的專利通信文件必須有人簽名，以證明其真實性。從業(yè)者必須親自簽署他們的文件，不允許他人代簽。商標通信文件也必須由從業(yè)者親自簽署或按照USPTO指南進行電子簽名。簽名方依據(jù)37 CFR 11.18(b)條款確認聲明是真實的，經(jīng)過合理查詢，并非出于不正當(dāng)目的。這確保了包括使用人工智能輔助起草的文件在內(nèi)的所有文件都經(jīng)過審查并且是準確的。未能遵守規(guī)定可能會導(dǎo)致制裁。參與USPTO程序的個人必須遵守誠實與善意義務(wù)，具體規(guī)定見37 CFR 11.303和37 CFR 1.56(a)。這一義務(wù)要求披露所有與專利性相關(guān)的信息，適用于與USPTO的所有溝通，不僅限于與專利審查員的互動。這包括提交請愿書和OA答復(fù)等互動。披露義務(wù)作為這一更廣泛義務(wù)的一部分，要求揭示所有已知的相關(guān)信息。它禁止欺詐和故意的不當(dāng)行為，確保USPTO程序的完整性和可靠專利的頒發(fā)。個人必須及時糾正任何錯誤。這一義務(wù)還涵蓋復(fù)審、PTAB程序和專利期限延長（37 CFR 1.555(a)和37 CFR 42.11），它還包括披露發(fā)明人、當(dāng)事人和從業(yè)者使用人工智能工具的信息。誠實與善意義務(wù)對于維護專利程序中的信任和效率至關(guān)重要。雖然沒有義務(wù)通知USPTO關(guān)于人工智能的使用，但如果人工智能的使用與專利性相關(guān)，從業(yè)者必須披露這種使用。特別是，如果人工智能起草權(quán)利要求或在專利申請中引入缺乏顯著人類貢獻的實施例，這必須通知專利局。USPTO不容忍在任何程序或IT系統(tǒng)訪問中的欺詐或故意的不當(dāng)行為，包括未經(jīng)授權(quán)的人工智能使用。違規(guī)行為可能導(dǎo)致嚴厲的處罰。在USPTO實踐中使用人工智能可能會無意中向第三方披露客戶的保密信息，包括技術(shù)細節(jié)。人工智能系統(tǒng)可能會保留輸入數(shù)據(jù)，存在進一步使用或違反37 CFR 11.106規(guī)定的保密義務(wù)的風(fēng)險。從業(yè)者必須確保人工智能工具和第三方服務(wù)維護客戶數(shù)據(jù)的保密性，特別是在外部存儲數(shù)據(jù)時。此外，如果服務(wù)器由非美國人士管理或訪問，使用人工智能可能會引發(fā)國家安全和出口管制問題。人工智能開發(fā)者的數(shù)據(jù)泄露也可能危及保密性。從業(yè)者在使用前必須審查人工智能工具的條款、隱私政策和網(wǎng)絡(luò)安全實踐。人工智能系統(tǒng)不被視為訪問USPTO電子系統(tǒng)的“用戶”。用戶必須創(chuàng)建單獨的USPTO.gov賬戶用于專利和商標申請，賬戶不得共享（人工智能無法創(chuàng)建賬戶）。專利中心、P-TACTS和商標電子申請系統(tǒng)需要注冊賬戶。此外，不得使用人工智能系統(tǒng)生成大量流量訪問USPTO數(shù)據(jù)庫，也不得以任何其他方式違反USPTO的服務(wù)條款。USPTO的指導(dǎo)意見澄清了人工智能發(fā)明權(quán)問題，確保了適當(dāng)?shù)男庞煤秃弦?guī)性。現(xiàn)有規(guī)則保護免受人工智能風(fēng)險的影響，允許使用人工智能輔助起草文件。除非有要求，否則通常不需要披露人工智能的使用。必須保持保密性、國家安全和適當(dāng)?shù)臄?shù)據(jù)使用。
英文稿
USPTOGuidance on AI-Related Patent and Trademark Issues
AI has introduced tools for content creation, legal research, and due diligence that can lower costs and improve service quality, but it also carries risks such as inaccurate outputs and confidentiality concerns. To address this, USPTO released the “Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office” on 4/11/2024 and the “Inventorship Guidance for AI-Assisted Inventions” on 2/13/2024 to inform individuals about AI-related risks, provide mitigation suggestions, clarify issues regarding inventorship, address confidentiality and national security concerns, and reinforce the duty of candor and good faith.Under 35 U.S.C. 115 and 116, inventorship is limited to natural persons under U.S. patent laws. The Patent Act defines an inventor as a human being, focusing on the mental act of conception, which AI cannot perform. However, AI can assist in inventions. This means that while AI systems cannot be listed as inventors on patents, their use does not disqualify natural persons who significantly contribute to the invention from being listed as such. This ensures the patent system incentivizes human ingenuity and provides protection for AI-assisted inventions with significant human contributions.The patent statutes require naming all inventors who contributed to at least one claim of a patent, focusing on those who contributed to the invention's conception. For AI-assisted inventions, each named inventor must make a “significant contribution,” which means they must significantly contribute to the conception or reduction to practice of the invention, make a meaningful contribution to the claimed invention, and go beyond merely explaining well-known concepts. Significant contribution is determined using the Pannu factors, which assess the inventor's role in conception or reduction to practice. This applies to single inventors using AI and joint inventors. Each claim in a patent must have at least one significant human contribution, and if an AI system is listed as an inventor, the application will be rejected. Maintaining “intellectual domination” over an AI system without significant contribution to the concept of the invention does not, on its own, make an individual an inventor of any inventions created using it.The USPTO provides guidance to determine if a natural person's contribution to an AI-assisted invention is "significant" for inventorship. Key principles include:1. AI Utilization: Simply using AI doesn't make someone an inventor unless they significantly shape the AI's output.2. Problem Recognition: Identifying a problem alone doesn't qualify as inventorship unless the person significantly guides the AI to a solution.3. Reduction to Practice: Merely reducing an invention to practice isn't enough; enhancing AI output meaningfully may qualify.4. Essential Building Blocks: Creating key components or designing AI to solve specific problems can be significant.5. Intellectual Domination: Control over AI doesn't confer inventorship without a substantial inventive contribution.Practitioners must avoid using AI-generated images as evidence of trademark use, as they don’t show actual use of a trademark and are therefore considered misstatements of fact. Additionally, AI evidence should be used with caution in US Trademark Trial and Appeal Board proceedings to avoid unintentionally submitting irrelevant information that may be thrown out as improper.Drafting Patent Documents for USPTO FilingComputer tools, including generative AI, are increasingly used in drafting documents for USPTO submissions, such as patent claims and Office action responses. While there is no prohibition against using AI for this purpose, all submissions must be reviewed and signed by the individual presenting the paper to certify its accuracy and compliance with USPTO regulations, specifically under 37 CFR 11.18(b). This includes verifying the accuracy of AI-generated content and ensuring it meets legal and evidentiary standards.All patent correspondence filed with the USPTO must bear a person's signature, certifying its authenticity. Practitioners must sign their documents themselves, not allowing others to sign for them. Trademark correspondence must also be personally signed by the practitioner or electronically signed following USPTO guidelines. By signing, the party certifies under 37 CFR 11.18(b) that statements are true, made with reasonable inquiry, and not for improper purposes. This ensures documents, including those drafted with AI assistance, have been reviewed and are accurate. Failure to comply may result in sanctions.Duty of Candor and Good FaithIndividuals involved in USPTO proceedings must adhere to a duty of candor and good faith, as detailed in 37 CFR 11.303 and 37 CFR 1.56(a). This duty requires disclosing all information material to patentability, extending to all dealings with the USPTO, not just patent examiners. This includes interactions such as filing petitions and responses. The duty of disclosure, a component of this broader duty, mandates revealing all known material information. It prohibits fraud and intentional misconduct, ensuring the integrity of USPTO proceedings and the issuance of reliable patents. Individuals must correct any errors promptly. This duty also covers reexaminations, PTAB proceedings, and patent term extensions (37 CFR 1.555(a) and 37 CFR 42.11). It includes disclosing information about the use of AI tools by inventors, parties, and practitioners. The duty of candor and good faith is essential for maintaining trust and efficiency in the patent process.While there is no general duty to inform the USPTO of AI use, practitioners must disclose AI usage if it is material to patentability. Specifically, if an AI system drafts claims or introduces embodiments in a patent application that lack significant human contribution, this must be reported. Patent protection requires a natural person's significant contribution, consistent with the USPTO’s Inventorship Guidance for AI-Assisted Inventions. The USPTO emphasizes the importance of internal documentation to comply with disclosure duties and counter future invalidation attempts. This stance aligns with the U.S. Copyright Office's position on human authorship for AI-generated materials.
The USPTO does not tolerate fraud or intentional misconduct, including unauthorized AI use, in any proceedings or IT system access. Violations may result in severe penalties.Using AI in USPTO practice can inadvertently disclose confidential client information, including technical details, to third parties. AI systems may retain input data, risking further use or breach of confidentiality obligations under 37 CFR 11.106. Practitioners must ensure AI tools and third-party services maintain client data confidentiality, especially when storing data externally. Additionally, AI use can raise national security and export control concerns if servers are located or accessed by non-U.S. persons. Data breaches by AI developers can also compromise confidentiality. Practitioners must review AI tool terms, privacy policies, and cybersecurity practices before use.USPTO’s Electronic Systems PoliciesAI systems are not considered “users” for the purpose of accessing USPTO’s electronic systems. Users must create individual USPTO.gov accounts for patent and trademark filings, which cannot be shared (AI cannot create one). Registered accounts are required for Patent Center, P-TACTS, and trademark electronic filing systems. Additionally, AI systems should not be used to generate high volumes of traffic to the USPTO database or in any other way violate USPTO’s terms of service.Conclusion:
The USPTO’s Guidance clarifies AI inventorship issues, ensuring proper credit and compliance. Existing rules protect against AI risks, and AI-assisted drafting is allowed. Disclosure of AI use is not generally required unless requested. Confidentiality, national security, and proper data use must be maintained.
作者：北京品源專利代理有限公司 Alexander Zeng
首頁
電話咨詢
聯(lián)系方式
