This article has been written by Nagesh H. Karale, pursuing a Diploma in US Intellectual Property Law and Paralegal Studies course from LawSikho, and edited by Koushik Chittella.

Introduction

Patent offices play a very important role in management of IP rights, which include patents, trademarks, registered designs, etc. Worldwide, patent offices promote innovation, protect IP rights, and drive economic growth for businesses and innovators. IP registration at patent offices gives SME’s the ability to succeed in innovation and survive in a competitive business environment. AI is continuing to evolve as an “assistive technology” rather than a replacement technology.

Functions of patent offices

There are various functions of a patent office, some of which are:

• Patent offices spread information about IP to the general public.
• They maintain IP ownership records and ensure accurate documentation.
• Patent offices oversee patent applications, grants, and maintenance.
• They play a role in enforcing IP rights and combating unauthorised use by addressing infringement and piracy.
• Patent offices help companies protect their confidential information, following Non-Disclosure policies.
• They assist businesses in staying updated on competitors’ IP awards, infringements, and claims.

Role of AI in patent processes

The role of AI in patent processes can be:

• AI algorithms analyse vast amounts of data (Patents/Trademarks, scientific papers) to identify prior art efficiently.
• AI assists in speeding up the drafting of patent applications.
• AI categorises patents / trademarks based on content to improve searchability.
• AI predicts patent trends to assist strategic decision-making for patent acquisition or portfolio management.
• Predictive analytics by AI tools assess patent value and licencing opportunities.
• AI identifies potential infringements by analysing existing patents/ trademarks.
• AI streamlines patent examination within offices.
• AI analyses legal texts, case law, and patent disputes to predict litigation outcomes.
• AI monitors online platforms for trademark infringement and copyright violations. Automated takedown requests and brand protection become more effective.
• AI analyses scientific literature, research papers, and industry trends. Companies can identify emerging technologies and potential collaborators.
• AI-powered platforms connect inventors, startups, and corporations to find solutions and foster collaboration across borders.
• AI-driven educational tools enhance IP literacy, which empowers inventors, entrepreneurs, and students with knowledge.

Evolution of patent processes

According to a study conducted by the Japan Patent Office in 2018, examiners without outsourcing at the JPO Patent Office usually spend:

1. 40% of their time is spent understanding the claimed inventions.
2. 30% in prior art search,
3. 10% in understanding the prior art documents found by the examiner, and 
4. 20% in evaluation of patentability and drafting first office action.

The necessary skills to execute the first three tasks require higher technical experience. The administrative action of the last task (fourth) and/or issuing a second examination report that contains another notification or a decision for grant or refusal must necessarily be issued by a public servant. Some national laws require an applicant to submit prior art documents known at the time of filing a patent application to assist the examiners of the patent office in conducting a substantive examination.

The substantive examination of the patentability requirements and the decision to grant or refuse a patent application should remain under the authority of the examiner. The success of the outsourcing programme will depend on the maintenance of the quality of service. In Brazil, both searches and examinations are made almost simultaneously by the same examiner. In EPO, the examination will also be done by the same examiner who made the search under the BEST programme (Bringing Examination and Search Together). The search report is transmitted to the applicant before publication.

Challenges faced due to traditional patent examination 

There are various challenges faced due to the traditional patent examination, which are:

• Prior art can include previously patented inventions, published articles, academic papers, technical documentation, product manuals, and even publicly available information on the internet. The sheer volume and diversity of prior art make the patent examination process complex and time-consuming.
• Patent examiners in different places and legal systems might assess patent applications differently. This happens because they each have their own understanding of patent laws. Additionally, they have varying levels of resources and different levels of expertise in different fields.
• Manual review methods used by patent offices cause delays in approving patents, which slows down product development and investment.
• Traditional methods of checking patents have trouble understanding all the tricky technical details in really complicated applications. This makes it tough to examine patents properly.

AI automates administrative tasks in processing patent applications, which facilitates managing paperwork and organising documents efficiently. It reduces the workload for patent clerks and lawyers. So, it allows them to focus on strategic aspects like claim construction and strategy development. It is important to make sure all the information in a patent application is correct. It helps prevent delays or rejections. This means double-checking details like applicant names and inventor information to keep the process accurate and reliable.

Challenges faced in managing large volumes of patent applications

In 2022, around USD 2.476 trillion was estimated to be invested in research and development (R&D) worldwide. This made many people apply for patents. Because of this, patent offices got very busy and couldn’t process patents quickly. This made inventors unhappy because it took too long to get their patents approved. Also, prior art searches take a lot of time and require special knowledge.

Slow patent processing causes legal uncertainties and stops innovation. It also makes companies not want to spend money on research and development (R&D), which reduces a country’s economic competitiveness.

Innovation and technological advancement in patent procedures

AI can compare information in patent applications with existing patents and other documents. It finds similarities that examiners can check when looking for earlier inventions. Using AI tools helps examiners go through applications faster for more than 70% of them.

Searching for earlier inventions, called “prior art search,” is hard and takes a lot of time. A study by the European Patent Office found that going through about 1.3 billion technical records across 179 databases gives about 600 million documents every month.

The Canadian Intellectual Property Office (CIPO) uses AI search engines to see connections between references, applications, and the latest technology. The Japan Patent Office (JPO) also uses AI to organise files. It gives relevant patent categories and words and ranks earlier inventions based on their importance. The United States Patent and Trademark Office (USPTO) uses AI to help decide if something can get a patent and to look at the history of patents.

INPI (National Institute of Industrial Property), the Brazilian patent office, uses AI tools and has made it quicker to examine and search for patents. It has reduced the time needed by up to 50%, or 75%, for some applications. This has also helped cut down the overall number of pending patents by 80%.

Role of artificial intelligence in patent offices

AI-assisted workflow integration leads to easy access and analysis of supporting documents. It tracks certain reference documents that are returned and how results are generated. These are useful for patent prosecution and internal quality reviews. AI is revolutionising patent workflows by automating tasks that accelerate patent filing and improve the accuracy of assessing existing patents and literature.

The “out-of-the-box” property of many AI models and techniques, when combined with large data, cloud computing, 5G, or the Internet of Things (IoT), enables AI to solve technical problems in almost any domain. It will also play a major role in identifying protein structures, drug interactions, and analysing RNA and DNA structures. AI predictive models are capable of forecasting patent trends, gauging application success, and even anticipating litigation probabilities. Additionally, AI aids in IP by spotting potential patent infringements

Google Cloud’s white paper introduces a methodology to train a BERT (bidirectional encoder representation from transformers) model on over 100 million patent publications, enabling various use cases. These include prior art searching, generating classification codes, and autocomplete features. It was released by Google in 2018. Its widespread adoption across domains like search and chatbots makes it a powerful tool. Patents pose unique challenges due to their large, complex, and context-dependent nature, making them ideal candidates for BERT’s capabilities to assist patent examination and spur innovation in the patent research community.

IP Australia uses the TMICS (Trade Mark International Classification Service) API for searching the Madrid Goods and Service (MGS) database, improving the quality of trademark applications when filed overseas. The Trade Mark Precedent Identification (TMPI) tool for Trade Mark Examiners uses various NLP techniques to enhance search quality and consistency. TM Checker is a free AI-assisted trademark availability check tool useful for educating small to medium enterprises about eligibility and potential distinctiveness issues for proposed trademarks.

Patsnap is a commercially available AI tool used for providing insights into markets, competitors, and partnership opportunities by analysing patents, scientific journals, and litigation. Questel, STN, and Clarivate Analytics are commercially available semantic AI search engines that assist patent offices in conducting searches for prior art and citations using machine learning algorithms.

PQAI (Patent Quality Artificial Intelligence), Ambercite, IPRally, Patseer, and InnovationQ Plus are a few examples of the AI-powered patent search databases.

Automation of patent search and examination

The European Patent Office formed a dedicated data science team with the goal of applying AI and ML technologies to increase the efficiency and quality of the patent grant process in 2019. These projects, focused on natural language processing, computer vision, and machine translation, are useful to the patent grant process in the areas of search, classification, and machine translation.

The five largest IP offices (the EPO, JPO, KIPO, CNIPA, and USPTO) receive over 80% of all patent applications in the world. These five offices are collectively known as the IP5. They cooperate on a variety of projects to improve and harmonise the patent system around the world. In 2019, the IP5 offices decided to advance their cooperation in the area of new emerging technologies and AI by setting up a special task force.

The IP5 partner offices and WIPO are currently exploring the impact on legal, technical, and policy aspects of new technologies and AI. The aim is to pinpoint areas such as joint IP5 responses by employing AI tools and systems support for patent examiners. They are also focusing on improving the patent grant process to apply the patentability requirements to inventions in the field of AI and AI-generated inventions.

Enhancing efficiency in patent prosecution

AI tools assist patent attorneys in drafting comprehensive responses to office actions by quickly analysing reference patents, which reduces the time and costs associated with prosecution. These tools can also generate first drafts based on similar rejections and responses. It helps applicants plan patent allowance applications by predicting success or advising on the likelihood of appeal.

AI initiatives in the European Patent Office (EPO)

1. Image search (trademark, design): EPO utilises ML and AI for patent searches, including automatic figure and image searches for patent drawings. EUIPO (European Union Intellectual Property Office) has created an in-house image search system integrated into eSearch Plus, enabling users to search for trademarks and designs using images. 
2. Patent prior art search: The EPO employs AI for patent searches, including automatic prior art search and query generation, using both in-house and commercial solutions.
3. Patent/Trademark classification: EPO employs AI for patent classification: pre-classification of incoming applications, document classification according to CPC, and re-classification for CPC updates.
   1. EUIPO (European Union Intellectual Property Office) utilises AI-based semantic search in Easy Filing to assist users in finding appropriate protection for their trademarks by enhancing the accuracy of goods and services searches.
   2. EUIPO has created AI-based tools to extract key information from letters and make decisions based on this data. These tools are used to analyse classification, formalities, and AG deficiencies in trademark applications, as well as to assess deficiency rates and grounds in design applications.
4. Patent examination management: The EPO employs ML and AI for patent file management, enhancing processes like annotation and problem detection through the Patent Document Model (PDM) in the Knowledge and Information Management Environment (KIME) for data enrichment. EUIPO has developed an algorithm to predict the outcomes of comparisons between pairs of goods and/or services using historical data and identifying semantically relevant matches. Currently, the tool is accessible only to examiners.
5. Machine translation: The EPO employs Patent Translate and is developing its own machine-learned translation tool for patent documents. This resource is accessible to the public via EPO patent databases and is utilised by trained examiners at other patent offices as well. EUIPO employs machine translation for Case Law documents via eSearch Case Law, offering automatic translations on its website for EUIPO decisions to help users understand the main content.
6. Data analysis: The EPO Data Science team uses open-source software and examiner knowledge to make AI systems. These systems look at trends in computer-implemented invention technology.
7. Helpdesk services: The EU Intellectual Property Office (EUIPO) now has its first chatbot, which is part of easy filing. It assists users with trademark questions by giving standard answers and can connect with a human agent if needed.

The US Patent and Trademark Office (USPTO) made an AI tool that automatically classifies patent applications using the CPC system. This saves the USPTO money in getting CPC data, and it also saves time and money for patent applicants.

Implications for patent office personnel

AI has made patent searches easier and quicker. Before AI, examiners had difficulty finding old patents because of the different words used. But now, AI looks at similar words and how they’re used to find patents better. It helps examiners quickly find old patents during reviews. AI also lets examiners focus on harder tasks. 

People who work in patent offices need to learn about AI and how it works. Training should be easy to access and affordable for everyone. We need clear rules to decide who owns the results that AI finds. It’s important to protect patent information and make sure AI is fair for everyone. These changes help make sure the patent system works well. AI builds trust with people when it’s honest about how it works.

Challenges and limitations of AI in patent processes

There are various challenges due to the usage of AI in patent processes, some of them are:

• Patent documents use special words and cover many years, which makes it hard to have all the information and keep it consistent.
• It is very difficult to get all the necessary information because people are concerned about their privacy. We must gather data in an ethical manner to properly train AI models, and collaborate closely with patent offices. 
• Teaching AI models to draft patents well requires understanding legal and technical terms, as well as expertise in particular areas.
• AI programmes might have biases based on the data they’re trained on, like gender or race biases. Since AI learns from provided data, it’s crucial to check that it’s fair and unbiased in its analysis.
• AI-created inventions bring up legal questions about who owns and invents them. Some countries accept AI-made inventions, but others demand humans be involved in making them.
• Due to the unique combination of legal and technical language present in patent documents, AI models require specialised training to accurately draft patents.
• Assessing patent validity and infringement using AI may require more resources.
• AI encounters challenges with technical and legal terminology, particularly when operating across different languages. There is a risk of AI misinterpreting patent information, which leads to errors in assessing patent validity, infringement, and licencing.
• AI models sometimes lack transparency, operating as “black boxes,” which can make it difficult to understand the rationale behind their decisions.
• Small patent offices in some areas struggle to use advanced AI systems because they don’t have enough patent data. Sharing data is the solution for this problem.
• AI systems might accidentally share private information and have biases, so we need to be careful. Laws like GDPR make sure that people’s privacy is protected.

Importance of human oversight and quality control

AI systems are good at tasks like recognising patterns and analysing data, but they can’t understand or make judgements like human experts can, especially when dealing with complex inventions or unusual prior art. So, it is important to see AI tools as supplements, but not as replacements for patent examiners. 

Human expertise is needed to guide AI systems and make sense of their results, which ensures that final patent decisions follow legal and ethical rules. Patent professionals provide important legal and contextual knowledge for AI-based patent analysis. Human intervention ensures AI decisions match human values. A mix of AI’s efficiency and human understanding is required to keep patent quality high.

Future directions and opportunities

Potential future developments and innovations for patent offices

EPO has launched the “Master the Prior Art” programme to improve classification procedures earlier in examinations. It increases search accuracy and retrieves relevant documents. The digital patent granting process integrates artificial intelligence, machine learning, and other technologies systematically.

In the digital patent granting process, AI-assisted solutions provide different functions. These include assigning patent classifications, tools to convert documents, APIs for easy document delivery, and online tools for reviewing and analysing patents. 

Advanced neural networks are improving the accuracy of extracting information from patent documents. AI-assisted patent searches might soon include text, images, and voice for a better search experience. Blockchain technology could ensure secure and transparent patent searches. Advanced AI algorithms can predict potential prior art, which helps spot issues early in the patent process.

AI algorithms could help with valuing patents and negotiating licences. The future of AI in finding patents seems bright. Quantum computing is bringing big changes to how patents are analysed. AI platforms help inventors, lawyers, and experts work together to speed up patent applications. Patent offices use AI to make reviews better and patents of higher quality. Patent searches benefit from AI’s improved language understanding, which leads to greater accuracy.

AI needs to get better at understanding images and diagrams in patents, which means we need to keep improving computer vision and deep learning. Tools that use AI to find patents should be transparent so that people trust them.

Opportunities for collaboration between patent offices, technology developers, and research institutions

Below are several promising opportunities for collaboration among patent offices, technology developers, and research institutions.

• Patent offices and research institutions can team up to share patent data for trend analysis. This data helps technology developers create AI models for predicting patent outcomes. It could be useful to identify infringements and assess patent quality.
• Patent offices and institutions can collaborate on research projects to develop AI algorithms for better prior art searches and to improve patent classification systems. They can also work on automating patent translation and document processing to enhance patent examination processes.
• Technology developers and research institutions can offer training programmes for patent examiners and office staff. These programmes include workshops, webinars, and knowledge-sharing sessions. The focus is on AI, data analytics, and emerging technologies.
• Patent offices can encourage jointly organised hackathons for technology developers. These events aim to create solutions for specific patent-related problems. They foster creativity and lead to novel tools for patent offices.
• Online platforms serve as a collaborative effort for patent offices, AI system developers, and AI researchers to share ideas and propose patent-related solutions.
• Guidelines for AI usage in patent offices can be collectively formulated to address concerns regarding bias, transparency, and fairness in AI algorithms.
• Technology incubators and accelerators can collaborate with patent offices to provide support for startups. They also offer mentorship, access to patent databases, and legal guidance.

Conclusion

AI can help patent offices handle more patent applications faster and more accurately. This means better management of intellectual property and more innovation. Patent-related AI systems face challenges including incomplete data, privacy concerns, and biases in AI algorithms. AI-driven systems improve patent search work and examination, streamline workflow, and help reduce backlog. It can also enhance patent quality, which leads to innovation and economic growth.

References

1. https://ip.com/blog/can-ai-invent-independently-how-ai-is-changing-the-patent-industry/
2. https://xlscout.ai/the-synergy-of-ai-and-patent-workflows-efficiency-accuracy-and-innovation
3. https://www.epo.org/en/news-events/in-focus/ict/artificial-intelligence
4. https://cloud.google.com/blog/products/ai-machine-learning/how-ai-improves-patent-analysis
5. https://www.epo.org/en/news-events/in-focus/ict/artificial-intelligence
6. https://www.wipo.int/about-ip/en/artificial_intelligence/search.jsp
7. https://powerpatent.com/blog/ais-impact-on-the-patent-examination-process
8. https://powerpatent.com/blog/patent-portfolio-cost-cutting-strategies-using-ai-assistance
9. https://powerpatent.com/blog/ai-assisted-patent-search-algorithms
10. https://www.wipo.int/wipo_magazine_digital/en/2023/article_0001.html