广东省贸易摩擦应对综合服务平台
（海外知识产权）

Comparing AI Patentability Across Jurisdictions

比较不同司法管辖区的人工智能可专利性

In each above-mentioned jurisdiction, AI inventions face unique requirements for patent eligibility, focusing on aspects like technical effect, application context, and reproducibility.

在上述各个司法管辖区，人工智能发明都面临着独特的专利资格要求，主要集中在技术效果、应用背景和可复制性等方面。

Brazil: Emphasizes that AI-based inventions must demonstrate a “technical effect” beyond standard calculations or data manipulation, focusing on concrete applications, particularly in industrial processes. Brazilian law excludes software “per se” from patentability, meaning that pure algorithms without technical integration are ineligible. For an AI invention to be patentable, it must solve a specific technical problem in a non-obvious and innovative way, as underscored by recent BPTO decisions.

巴西：强调基于人工智能的发明必须证明超出标准计算或数据操作的 “技术效果”，重点关注具体应用，尤其是在工业流程中的应用。巴西法律将软件 “本身 ”排除在专利之外，这意味着没有技术整合的纯算法不符合专利申请条件。人工智能发明要想获得专利权，必须以非显而易见的创新方式解决具体的技术问题，这一点在巴西专利局最近的裁决中得到了强调。

United States: Follows a “two-prong” approach, particularly under recent USPTO guidelines. Prong One assesses whether an invention falls into an abstract category, such as mathematical methods or mental processes, which could make it ineligible. Prong Two examines if the claim incorporates a practical application in a technical field, ensuring that it delivers tangible improvements. The USPTO’s framework is geared towards granting patents only to AI technologies with a concrete technical impact, such as enhancements in network security or medical diagnostics, thus discouraging abstract claims without functional applications.

美国：采用 “双管齐下 ”的方法，特别是根据美国专利商标局最近的指导方针。第一要素评估发明是否属于抽象类别，如数学方法或心理过程，这可能使其不符合资格。第二步审查权利要求是否包含技术领域的实际应用，确保其带来切实的改进。美国专利商标局的框架只向具有具体技术影响的人工智能技术授予专利，如增强网络安全或医疗诊断，因此不鼓励没有功能应用的抽象权利要求。

Europe: The EPO also adopts a technically grounded approach, but with an added requirement for a “specific technical purpose”. The EPO guidelines emphasize the need for detailed disclosures, including technical aspects of algorithms and machine learning models, which must be presented in a way that ensures reproducibility. This approach requires inventors to describe the nature of data and technical structure of AI models comprehensively, ensuring that a skilled person can replicate the invention without undue experimentation. Similar to Brazil, software or algorithmic inventions that are not directly tied to a practical technical application are unlikely to qualify as patentable.

欧洲：欧洲专利局也采用以技术为基础的方法，但增加了 “具体技术目的 ”的要求。欧洲专利局的指导方针强调了详细披露的必要性，包括算法和机器学习模型的技术方面，必须以确保可重复性的方式进行展示。这种方法要求发明人全面描述人工智能模型的数据性质和技术结构，确保技术熟练的人可以复制发明，而无需进行不适当的实验。与巴西类似，与实际技术应用没有直接联系的软件或算法发明不太可能获得专利。

Focus on Inventive Step and Reproducibility

注重发明步骤和可复制性

While all three jurisdictions require inventive step, their approaches differ slightly:

虽然这三个司法管辖区都要求发明步骤，但其方法略有不同：

In Brazil, the inventive step is strongly linked to the technical effect that the AI delivers. The AI solution must address a technical problem in a novel and innovative way, emphasizing the invention’s applicability in industrial and commercial settings.

在巴西，创造性与人工智能带来的技术效果密切相关。人工智能解决方案必须以新颖和创新的方式解决技术问题，强调发明在工业和商业环境中的适用性。

The USPTO examines the inventive step in conjunction with the “two-prong” approach, where only those AI aspects that contribute to the technical purpose are considered. The USPTO’s recent emphasis on technical impact further filters for inventions that contribute to specific technological advancements.

美国专利商标局结合 “双管齐下 ”的方法审查创造性步骤，即只考虑有助于实现技术目的的人工智能方面。美国专利商标局最近对技术影响的强调进一步过滤了有助于具体技术进步的发明。

The EPO places special emphasis on reproducibility and technical specificity. By requiring that AI inventions serve a “specific technical purpose,” the EPO focuses on inventions that achieve tangible technical outcomes. Additionally, EPO guidelines demand detailed descriptions of the technical workings and characteristics of data involved in AI training processes, enabling a person skilled in the art to replicate the invention effectively.

欧洲专利局特别强调可复制性和技术特殊性。通过要求人工智能发明服务于 “特定技术目的”，欧洲专利局将重点放在能够实现有形技术成果的发明上。此外，欧洲专利局的指导方针要求详细描述人工智能训练过程中涉及的技术工作原理和数据特征，使本领域的技术人员能够有效地复制发明。

Stick to Technical Solutions

坚持技术解决方案

Brazil, the United States and Europe have each adopted stringent requirements for patenting AI inventions, tailored to promote technological advancement while preventing overly broad, abstract claims. Brazil and Europe align closely in focusing on the technical effect and industrial applicability of AI, with Europe adding a rigorous requirement for reproducibility and technical specificity in algorithmic descriptions. The United States, through its “two-prong” test, emphasizes filtering out abstract AI inventions unless they are integrated into practical applications with technical benefits.

巴西、美国和欧洲都对人工智能发明专利采用了严格的要求，旨在促进技术进步，同时防止过于宽泛、抽象的权利要求。巴西和欧洲在关注人工智能的技术效果和工业适用性方面密切配合，欧洲在算法描述中增加了对可重复性和技术具体性的严格要求。美国则通过 “双管齐下 ”的检测，强调过滤抽象的人工智能发明，除非它们被整合到具有技术效益的实际应用中。

In conclusion, while each jurisdiction maintains distinct requirements, the underlying goal is similar: to support innovation in AI that provides concrete technical solutions and contributes meaningfully to the industry. Inventors seeking patents across these regions should focus on demonstrating how their AI technologies solve specific technical problems, deliver reproducible technical improvements, and contribute to defined technical fields.

总之，尽管各个司法管辖区有不同的要求，但其根本目标是相似的：支持在人工智能领域提供具体技术解决方案并对行业做出实质性贡献的创新。寻求在这些地区获得专利的发明人应重点展示其人工智能技术如何解决特定的技术问题，带来可再现的技术改进，并为特定技术领域作出贡献。