Analysis of U.S. Policies for Promoting AI Patents
Pine IP
February 12, 2025
.avif)
As artificial intelligence (AI) technology has rapidly advanced in recent years, the U.S. government and the United States Patent and Trademark Office (USPTO) have been promoting various policies to foster AI patents and refine the intellectual property framework. In particular, these policies are focused on strengthening institutional support to encourage companies and research institutions to actively innovate in the AI field and to ensure their creations are securely protected. Below, we will examine the current status of U.S. AI patent policy, its impact, related legal and ethical issues, international comparisons, and strategic implications for businesses.
The USPTO is continuously refining its policies and guidelines to encourage inventions in the field of AI. Since the 2020s, AI-related patent applications have surged, with annual filings more than doubling since 2002 and increasing by over 33% since 2018. In response, the USPTO announced its “2025 USPTO AI Strategy” in January, presenting a vision to promote and responsibly protect AI innovation. This strategy includes key initiatives such as fostering inclusive AI innovation, strengthening the USPTO's internal AI capabilities, encouraging the responsible use of AI, cultivating professional talent, and promoting domestic and international cooperation. It sets a policy direction for the U.S. to maintain its leadership in patents during the AI era.
In terms of patent examination, guidelines for AI-related inventions have become more specific. In February 2024, the “Inventorship Guidance for AI-Assisted Inventions” was published, clarifying that an invention involving AI can be patented if a human inventor has made a substantial contribution. It also emphasized that merely inputting a problem into an AI and receiving a solution is not sufficient to qualify as a human inventor, ultimately reaffirming the principle that under patent law, an inventor must be a human being (as established in Thaler v. Vidal by the U.S. Court of Appeals for the Federal Circuit). While AI itself cannot be named an inventor, this guidance makes it clear that inventions created by humans using AI as a tool are eligible for patent protection, thereby actively encouraging R&D that utilizes AI.
Furthermore, in July 2024, guidelines concerning the subject matter eligibility of AI inventions were updated to provide more detailed criteria for patent examination. The guidelines specify that under 35 U.S.C. § 101 and the Alice framework, an AI-related invention can be patented if it is tied to a specific technical application or improvement, rather than being an abstract idea. This indicates that while simply listing AI algorithms or models may be rejected as abstract concepts, a patent can be granted if the invention clearly presents a solution to a specific industrial application or technical problem.
At the government level, various forms of support for encouraging AI patents continue. The Biden administration's October 2023 Executive Order on AI directed the USPTO to develop further guidance on AI and patent issues, leading to enhanced measures regarding inventorship and patent eligibility. Since 2022, the USPTO has been running the AI/ET Partnership program to gather diverse opinions from industry and academia and has held public listening sessions to expand stakeholder participation in policy development. Additionally, programs like the Track One Prioritized Examination help innovative startups and researchers obtain patents more quickly. Legislation such as the National AI Initiative Act and the CHIPS and Science Act also promotes investment in AI R&D and infrastructure, which is expected to create a positive foundation for generating AI-related achievements through patent filings in the long term.
The USPTO's 2025 AI Strategy report outlines five key AI Focus Areas:
U.S. policies promoting AI patents have a direct ripple effect on the innovation activities of businesses and research institutions. First, the USPTO's clearer examination guidelines provide AI patent applicants with greater predictability and transparency. For example, when developing new AI algorithms or models, companies can increase their chances of obtaining a patent by preparing documentation in advance that proves the contribution of human inventors and detailing it in their applications. This provides a strong incentive for companies to invest confidently in AI R&D, assuaging previous uncertainties about whether AI-driven research outcomes could be protected by patents.
Furthermore, the USPTO's criteria for the patent eligibility of AI inventions help companies systematically understand which technological elements of AI should be emphasized to secure a patent. Leading U.S. companies are rapidly expanding their AI patent portfolios. As of 2023, IBM leads with over 7,000 AI-related patents, while Google holds nearly 5,000. This is interpreted as a strategy for companies to protect their AI technological capabilities with intellectual property rights, establish a market advantage, and create a favorable environment for attracting investment and licensing technology.
For startups, in particular, securing a U.S. patent serves as an indicator of their technological prowess and exclusive rights, positively impacting subsequent investment rounds and corporate valuation. For instance, if an autonomous driving AI startup obtains a U.S. patent for its core algorithm, it can significantly enhance its negotiating power in partnerships or M&A deals with larger corporations. Research institutions and universities are also experiencing the benefits of these pro-patent policies. When AI technology developed by a university research lab is protected by a patent, it is more easily transferred to businesses or leads to spin-off ventures, a process actively supported by government programs. Ultimately, the combination of a clear patent protection framework and government support is strengthening the virtuous cycle of AI R&D → patent application → commercialization in both corporate and academic sectors.
The 2024 USPTO guidance, which clarifies that AI-assisted inventions can be patented if a human inventor has made a “significant contribution,” offers a balanced approach to inventions created through human-AI collaboration. However, the specific determination of what constitutes a "significant contribution" may vary on a case-by-case basis, leaving room for legal disputes. To prepare for this, companies are adopting strategies to document the ideation and AI utilization processes and to prove the involvement of human inventors.
The scope of patent protection is another critical issue. AI algorithms and mathematical models face a high risk of being considered abstract ideas. Since the Alice decision, the requirement for software and AI-related inventions to demonstrate a specific technical effect to be patentable has been strengthened. For example, if you invent a new machine learning algorithm, you are more likely to obtain a patent by claiming not just the algorithm itself, but also the industrial or technical improvements achieved by applying it. This encourages AI inventors to define their inventions from a practical application perspective.
IP infringement and licensing issues are also becoming more complex in the AI domain. Since AI technology often consists of various modules and algorithms, a single product or service can be covered by multiple overlapping patents. Consequently, companies are preemptively conducting Freedom to Operate (FTO) analyses to ensure their AI technology does not infringe on others' patents and are entering into license agreements when necessary to reduce the risk of disputes. Furthermore, the legal and policy interpretations regarding liability when an AI model's output infringes on someone else's patent are still unclear. For instance, if an AI generates a patented design or circuit, questions arise as to whether the developer or the user is liable for infringement when that design is incorporated into a product. In response, the USPTO is publicly gathering opinions on how AI impacts determinations of prior art, the "person having ordinary skill in the art" standard, and non-obviousness assessments to shape future legal principles.
On the licensing front, both collaboration and conflict coexist around AI patents. Major big tech companies tend to cross-license their vast AI patent portfolios to reduce disputes and maximize technology utilization. For example, companies like IBM, Microsoft, and Google, while holding large numbers of AI patents, also contribute to standardization bodies and open-source communities, sometimes disclosing certain patents or licensing them under specific conditions. However, AI patent disputes could become a new battleground for patent trolls or competitors. This connects to the ethical debate that "excessive patent monopolies can stifle innovation." Some research communities argue that core algorithms and models should be openly shared to accelerate progress, fueling an ongoing discussion about the balance between exclusive patent protection and technological innovation.
When comparing U.S. AI patent policy internationally, distinct differences emerge with major players like China and Europe. China leads the world in terms of patent application volume. According to some statistics, the number of generative AI inventions filed in China from 2014 to 2023 was about six times that of the U.S. In 2023 alone, China accounted for a significant portion of the world's generative AI patents, with the Chinese government promoting domestic patent acquisition as a national strategy through strong AI development policies. In the past, some local governments even offered subsidies for patent filings to encourage quantitative growth. As a result, in 2022 alone, approximately 40,000 AI patents were granted to Chinese inventors, a stark contrast to about 9,000 in the U.S.
However, despite China's quantitative dominance, there are concerns about the quality of its patents. According to NSF statistics, the average number of citations for Chinese AI patents is much lower than for U.S. AI patents. This suggests that patents from Chinese companies are primarily utilized and cited within China and have limited international impact. With the exception of a few large corporations like Huawei and Tencent, many Chinese companies tend to focus on securing domestic patents rather than filing internationally. In contrast, U.S. companies are active in securing global rights through PCT international applications in addition to U.S. patents. Global corporations like Samsung Electronics and Intel also pursue international IP strategies by obtaining patents in multiple countries.
Europe maintains relatively strict and consistent standards in its AI patent policy, reaffirming the principle that "an inventor of an AI invention must be human." The European Patent Office (EPO) primarily assesses whether an AI-related invention has a "specific technical purpose or technical implementation." While pure algorithms or mathematical methods are not patentable, inventions demonstrating a technical effect, such as a specific AI method for processing image data to remove noise, are patentable. This is similar to the U.S. abstract idea analysis, but Europe is stricter in requiring that the AI algorithm and training data be described sufficiently in the specification to be reproducible. Concurrently, the European Union is leading discussions on AI regulation, ethics, and safety through initiatives like the AI Act. In terms of patent disputes, while Europe is more limited than the U.S. in damages and procedures, the introduction of the Unified Patent Court (UPC) in 2023 has enabled unified patent protection across Europe, attracting growing interest from global companies.
Additionally, Japan and South Korea are steadily expanding their influence in the AI patent field. As of 2024, South Korea and Japan are filing over 1,500 AI patents annually each. Japanese patents are also considered to be of high quality, as their citation metrics are higher than China's. Germany also shows a high average citation count for its AI patents, with traditional manufacturing giants like Siemens and Bosch strengthening their patent strategies centered on industrial applications by integrating AI technology.
In this international comparison, the strength of the United States lies in its stable patent law system and enforcement capabilities. Rights enforcement through patent litigation is active, and large damage awards from jury trials are possible, making U.S. patents a valuable tool for business protection. In contrast, while patent registration in China is fast, there is an opinion that foreign companies still face challenges in effectively enforcing their patents there. Therefore, global companies are increasingly adopting IP strategies that consider the unique institutional characteristics, strengths, and weaknesses of the U.S., China, and Europe. A typical "two-track strategy" involves actively securing patents locally for technologies targeting the Chinese market, while also obtaining patents for core source technologies in the U.S. and Europe to establish international rights. A nuanced approach is required, such as tailoring specifications and claims to meet the different patent examination standards of each region (e.g., the U.S. Alice framework vs. Europe's technical effect requirement).
To fully leverage the U.S. policy environment for promoting AI patents, we recommend that AI companies consider the following strategies:
U.S. policies for promoting AI patents, through clearer regulations and government support, offer distinct opportunities for innovators in the AI space. This fosters confidence among companies and researchers that they can develop AI technologies more boldly and protect their outcomes with intellectual property rights, thereby accelerating investment in AI R&D. However, as new legal and ethical issues emerge, a prudent approach to managing them is necessary.
Only companies that closely compare the patent systems of major global players, including the U.S., and build patent strategies tailored to different regional examination standards will be able to lead in the AI technology and IP competition. In particular, building a strong patent portfolio in the U.S. market will become a powerful asset and a defensive shield in the future race for AI supremacy. From this perspective, we urge you to actively utilize AI patent policy to maximize your company's capabilities.
References:
