On February 26, 2022, Asia America Innovation Alliance (AAIA) hosted an online webinar to take a deep dive look into some critical issues and strategies in patenting AI. Two honored speakers Yanbin Xu and Yi Yu, along with the event moderator, Li Feng shared their experience and thoughts.
Introduction of Patents and AI
What is a patent?
A patent is a limited-time monopoly for an invention to prevent others from practicing that invention. It includes specification and drawings to disclose invention, and claims to define the scope of protection. A technology can be patented only if it is useful, novel, non-obvious, and more than an abstract idea. Both the disclosure and claim determine the scope of protection. Patent is considered as the key competitive advantage for a company and can drive significant business growth.
In the U.S., a patent application contains three steps: Internal Invention Disclosure, Patent Drafting, and patent examination. In the first step, the inventors, typically employed in a company will disclose their inventions through internal standardized forms, I.e. invention disclosure form. The legal department will draft the disclosure and claims internally or with external attorneys. Then the patent application will move to its last stage in the process for examination by the patent examiners. The patent application process ( “prosecution”) involves back-and-forth communication with the Patent Office, including “Office Actions” on the merits, responses to Office Actions, Amendments, Examiner Interviews, Appeals, etc. After a few rounds of Office Action, the applicant will be granted a patent right if the all requirements are met.
Patents Around the World
Patent is limited to national jurisdictions. If protection is needed in additional countries, the applicants should apply for patents in each country. Typically, the patent application can go through either Patent Cooperation Treaty (PCT) or Paris Convention. The former is more economical, while the later is faster in obtaining national patent protections.
Typically, Processes (e.g., methods), machines (e.g., devices and systems), compositions of matter (e.g., chemicals), as well as articles of manufacture (e.g. fabrics and plastics) can be patented. One cannot patent abstract ideas (e.g., business methods; mental processes), laws of nature (e.g., gravity) or natural phenomena (e.g., bacteria).
Patenting AI: Critical Issues
According to the U.S. National Institute of Standards and Technology (NIST) definition (2019) “AI technologies and systems “comprise software and/or hardware that can learn to solve complex problems, make predictions or undertake tasks that require human-like sensing (such as vision, speech, and touch), perception, cognition, planning, learning, communication, or physical action.”
Starting from the machine learning in the 80s, AI has evolved to Deep Learning. It can be used in a broad range of applications including speech, vision, planning/control, as well as knowledge processing.
In most of the countries (U.S., U.K, and EU), only a natural person can be considered as an inventor, therefore AI system is not eligible to be considered as inventor. Australia and South Africa would allow AI system to be inventors.
Strategies for Patenting AI
It is important for inventors to work with attorneys and business partners to identify strategies to improve patent’s eligibility. There are a few domains for consideration:
- Focus on Technical Aspects of Invention
- Use Language Rooted in Technology
- Avoid terminology that reads on mathematical concepts, mental thoughts or other excluded areas
- Avoid treating AI as a Black Box
- Look for human inventors to name in patent application
Furthermore, it is important to consider the following questions as the inventor and legal works on the claims:
Who infringes on the patent, competitor or clients? The patent may not have substantial value if it is likely the clients who infringe on the patent.
What is the scope of infringement? It is important to determine where the system or apparatus/component should be protected.
Where does infringement occur? It is important to determine where infringement occurs, and obtain the patent protection in said country.
Strength of the Patent. It is important to consider all the alternatives and include appropriate ones in the patent.
What is more valuable/inventive? An AI project is often large in scope. It is important to determine the value of each components (Data set for training the system or the AI system architecture/hardware). It is also worth to consider other forms of protection (e.g., trade secrets)
How can I detect/prove infringement? It is important to include claims directed to detectable features to make infringement more easily detectable.
AI patent filings have increased significantly over recent years and the trend is expected to continue. There are limitations to patent eligible subject matter in the U.S. Generally, the claims should focus on technological improvements over the prior art or a “technical effect.” Presently, an inventor is limited to a “natural person” and does not cover AI machines. For AI inventions, more specific details may be required to adequately disclose. AI should not be treated as a “black box”. The claims for an AI invention should be drafted with an eye towards enforcement. Consider issues such as who infringes, what infringes, and where infringement occurs to identify the most valuable claim types. Pursue as many possible claim types as possible for an AI invention. The invention may not be limited to your field of industry or require specific AI components to be patentable.