How the EPO and USPTO use and review AI

The USPTO and EPO are integrating AI and machine learning into their examination systems to help streamline the patent application process from start to finish, say sources at the EPO and USPTO; including everything from translation to data analysis and image searches.

One area where artificial intelligence (AI) will become an integral part of the system is prior art research. As the volume of prior art grows exponentially, AI and machine learning tools will be developed to help examiners review the landscape and make sure patent applications meet the necessary requirements for novelty.

Andrei Iancu, undersecretary for commerce and IP and director of the USPTO, says: “More advanced AI capabilities will assist examiners in sifting through this ever growing body of prior art. We believe we can leverage AI capabilities for classification, as well as other office activities.”

He anticipates that AI and machine learning tools will provide an additional benefit to the USPTO during the examination process by reducing administrative burdens on examiners and supervisors.

The EPO is likewise looking to AI to improve the quality of their research in the patent grant process, especially in the field of prior art. Jana Mittermaier, EPO spokesperson and director of external communication in Munich, says examiners will soon have ‘virtual assistants’ that will extract information from prior art data bases, and also learn from the examiners themselves.

The deliverables from these projects will be integrated into an electronic patent granting process. “The patent office favours a gradual approach, the idea being to incorporate AI into the patent grant process, not as a big bang, but as pilots first, extending the number of examiners that use them over time as the products are enhanced,” says Mittermaier.

“We have made digital transformation one of our key goals under the Strategic Plan 2023 for the next four to five years. The office will thus continue to invest heavily in the area of IT to contribute to higher quality and improved timeliness of our products and services, as well as increasing staff and user satisfaction.”

The EPO receives applications in several languages, and since 2017 it has used artificial neural network technology to translate patent dossiers. Their Patent Translate tool is available to the public and used by specially trained patent examiners. The UKIPO and the Swedish Patent and Registration Office also use the Patent Translate tool to facilitate patent application translations.

One area of concern for patent examiners, but not necessarily filers, is the fear that one day robots will take over their jobs. Mittermaier says she would like to alleviate these fears. “We foresee AI tools providing increasingly important support to assist the work of the highly skilled staff in patent offices across the world, but not to replace the officers,” she says.

Instead of replacing patent examiners, AI is doing more to bring IP offices around the world together. The Image Search Project is a cooperative effort of the five largest trademark offices in the world to enhance image search database systems.

“We also put great value on our co-operation with the national IP offices in Europe and beyond. That is why we are looking at building and implementing common tools with the national offices of our 38 member states. With our IP5 partners – the patent offices of China, Japan, Korea and US – there are ongoing exchanges of information about the use of AI among our experts,” says Mittermaier.

The USPTO Manual for Patent Examination Procedure states: “The requirement that the applicant for a patent be the inventor is a fundamental characteristic of US patent law.”

Iancu admits that the USPTO might find itself facing more applications with non-human inventors as AI becomes more developed, which will raise questions such as subject matter eligibility and ownership. In response to these concerns, the office has created a special AI task force to address these questions. 

Even if AI isn’t doing the inventing itself just yet, more patent applications are being filed in the area of AI and machine learning. According to the WIPO’s 2019 Artificial Intelligence Publication, AI published applications have grown by 400% in the past decade. 

The USPTO has seen a 34% increase in the share of AI patent filings since 2005 and has doubled the number of examiners it uses to review AI applications. New AI technologies make up about 26% of annual patent filings.

Europe is experiencing similar trends in AI patent filings. “We have also seen that AI-related patent applications have surged in recent years,” says Mittermaier. “In November 2018 we published an update to our Guidelines for Examination at the EPO which provides more specific guidance on the examination of AI applications.”

Like the US, a requirement for the grant of a European patent is that applicants must specify the names of the inventors. While there is no definition of inventor in European patent law, it is generally assumed that a human being is the creator of any invention.

“The inventor must be able to employ human faculties rather than merely produce a certain output. The current state of technological development suggests that, for the foreseeable future, AI is rather a tool used by a human inventor in the inventive process,” says Mittermaier.

Whichever way AI develops in the future, it is clear that it will help patent offices by making the jobs of examiner’s a lot easier, and also increase their work load by giving them more patent applications to review.