Within a couple decades or so advanced Artificial Intelligence (AI) may cause mankind’s next revolution. In the meantime, AI is pervasively impacting numerous areas of technology and the economy. As businesses rush to exploit AI, protection of IP is a key concern – funding, revenue, partnerships, and exits can depend upon it.
China, which already has advantages in certain aspects of AI1, recently released an AI development plan that features the strengthening of patent protection for AI. In the U.S., patent-eligibility is restricted for inventions that involve abstract ideas – and that restriction can impact AI-related inventions2. Vigilance regarding the legal landscape, if not keen foresight as to the near future, is needed as the U.S.-patentability of AI-related inventions is still evolving3. At a minimum, a skilled patent attorney should creatively assess potentially eligible AI-related inventions, strategically prepare a patent application that explains and claims inventions in patent-eligible terms, and be ready to help nudge a company’s R&D efforts toward patent-eligible areas.
At the same time, trade secret protection should be weighed and utilized in concert with patent protection. For aspects of an AI-related system or service that a company can keep secret confidently for the relevant period of time (years or more), it may be best to take measures to retain secrecy rather than disclose the matter by applying for a patent. It may also be possible to retain some aspects as trade secrets while seeking patents on others. A key issue in deciding between patent and trade secret is the feasibility of proving infringement – if there is not likely to be evidence of the infringing use of your AI-related invention, secrecy is likely the better course.
All personnel and entities that may create AI-related IP subject matter on your behalf should be under contractual obligation to assign their inventions and other IP to you before commencing that work.
If open source or other third party-licensed matter is utilized, all licensing terms should be reviewed closely at the outset for potential restrictions on patenting and other IP protection.
Other Patent Issues Commonly Relevant To AI
If patent protection is sought, there are a few other issues that can be particularly relevant to AI-inventions. First, the patent claims should be structured not only in light of the eligibility issue mentioned above, but also to: 1) exclude matter subject to third party ownership or restrictions; 2) focus on things likely to be infringed by a single party rather than distributed over multiple entities; and 3) focus on things as to which infringement is likely to be provable. The application and its disclosure likewise should be written to tell a compelling story thematically in light of the invention as it is actually being claimed.
Another legal requirement that can be relevant to AI-related inventions is “enablement.” Algorithms and data structures should be disclosed, along with detailed example(s) and results of training on external data, adequately so that one of ordinary skill in the field could carry out the claimed invention after reading the patent application on the day of its filing. To the extent particular details would not be necessary to meet that standard or to carry out the claimed invention’s “best mode,” however, such details can be left out of the patent application if it may be preferred to retain them as trade secrets.
Patenting Things “Created By” AI
Finally, there is a good bit of discussion by scholars and other commentators about scenarios in which AI creates inventions without human input. My view is that such hypotheticals are many years premature and based on a misunderstanding of the nature of today’s AI algorithms. Until AI has reached the “inflection” where it acquires something akin to genuine sentience, no AI is being creative or creating anything in the sense of “inventing.” Today’s AIs simply act as tools – no matter how unpredictable their results or product. Throughout the ages inventions have often been the fruit of unexpected results yielded by human application of human-made tools. The humans who devise AI algorithms, feed them data, and recognize inventive results are legally the inventors of that subject matter. AI is more powerful and can be more opaque than yesterday’s tools, yet for now it remains a tool devised and applied by humans.
- Notably, China has the most AI startups and half of the world’s AI venture funding, a larger population and deeper reservoirs of data, more centralization and less privacy regulation, and for a number of years it has had many times more AI patent application filings than the U.S.
- Most notable are Alice Corp. v. CLS Bank International, 573 U.S. 208, 134 S. Ct. 2347 (2014), which generally heightened the standards for patentability, and Electric Power Group, LLC v. Alstom SA, 830 F. 3D 1350 (Fed. Cir. 2016), which dealt with an AI-related invention.
- Subsequent court decisions could change course from Electric Power Group, and comments at a Senate oversight hearing last year suggest that patent-friendly changes may be coming from the U.S. Patent Office and/or Congress. See www.multichannel.com/news/ptos-iancu-ai-algorithms-generally-patentable.
Please complete the form below and we will contact you shortly.
All fields marked with a * are required.