For a small company, even the cost of applying for and maintaining a small number of patents is significant and therefore it is important that cost is wisely incurred. David Brinck, Partner and UK and European Patent Attorney at EIP, explains the reasoning behind small tech start-ups deciding whether to patent their Machine Learning (ML) algorithms.
The amount of product innovations relying on Artificial Intelligence (AI) and Machine Learning (ML) algorithms is constantly rising, with online platform, Crunchbase, reporting that the number of start-ups on its platform relying on it grew by 6% last year alone. As all these start-ups enter the market, there is a growing demand for advice and support regarding whether to file for patent protection.
There is an inherent tension between wanting to get a product out to market quickly before it loses its unique selling point and going through what can be a time-consuming process of filing for a patent to protect the product as an asset long-term. This balance act will only become more common and is something any start-up looking to launch an AI algorithm-based product cannot afford to ignore, even if it is daunting or confusing to begin with. Gigantic multinational companies with portfolios of many thousands of patents can have entire departments of staff dedicated to portfolio management. For a small company, even the cost of applying for and maintaining a small number of patents is significant and therefore it is important that cost is wisely incurred.
AI and patentability
As with many emerging technologies, the pace of development of AI tends to be faster than the development of the laws regulating it. Over the past few years, there has been an increasing debate over whether existing patent law is fit for AI-based innovations. Some argue that existing patent law is too strict and that patents should be more easily granted for AI-based innovation to encourage investment and commercialisation. Others argue that patents should not be granted for AI-based innovation as it has the potential to stifle innovation in a fast-moving field.
Patent Offices around the world are currently investigating what AI-based innovations should and should not be susceptible to patent protection. Given the on-going efforts to harmonise patent law throughout the world, these investigations will not be carried out in isolation and we can expect there to be broad conformity in the eventual outcome. Nevertheless, we can also expect there to be differences in the detail.
At the European Patent Office (EPO), a key question is whether there is genuine ‘technical’ innovation. In this regard, running a purely mathematical algorithm (such as an AI algorithm) on a computer is not in itself generally considered to be technically innovative in Europe. Two exceptions are (1) if the AI algorithm is being executed in the context of a technical application, (2) if the AI algorithm is implemented on a computer in such a manner that it addresses a problem within the computer, for example an excessive requirement for computational resources.
In contrast, at the US Patent and Trademark Office (USPTO) a key question is whether the AI-based innovation involves more than just an ‘abstract idea’. US patent examiners have often considered AI-related inventions to be no more than abstract ideas and therefore ineligible for patent protection.
In short, whether an AI-based innovation is likely to be patentable depends on one or both of the nature of the data being processed – data corresponding to objective measurements suggesting possible patentable subject matter – and whether there is any interaction with hardware going beyond a general purpose computer. While the practice at the patent offices is evolving, which types of technological applications are more susceptible to patent protection are being continuously monitored by patent professionals.
Taking the leap
Even for AI-based technology innovation that meets the criteria for patentability, in practice getting a patent granted can be challenging. One challenge is ensuring that the technical information provided with the patent application describes in sufficient detail how the invention works, while also discussing sufficient possible modifications to provide adequate protection from a commercial viewpoint. Another challenge is, put simply, persuading a patent office examiner that it is worthwhile to move forward with an application.
The value of a patent is intrinsically linked to the value of the underlying technology. If the underlying technology provides a commercial advantage, then protecting that underlying technology with a patent will have commercial value. There are, however, several other considerations that need to be borne in mind when deciding whether to apply for patent protection, including:
• Are you looking to collaborate with other parties? If you are, then this would be a factor in favour of applying for patent protection
• Are you looking for investment? If you are, then this would be a factor in favour of applying for patent protection
• Are there alternatives to the underlying technology that perform just as well? If there are, then this would be a factor against applying for patent protection
• How quickly is the underlying technology likely to become redundant? If it is likely to become redundant in a year or two, then this would be a factor against applying for patent protection
• Is it possible to keep the underlying technology confidential? If it is, then this would be a factor against applying for patent protection
• Similarly, could a competitor keep their use of the underlying technology confidential? If it is, then that would be a factor against applying for patent protection.
• Could there be a tax advantage to obtaining patent protection, such as the patent box scheme in the UK? If there could, then that would be a factor in favour of applying for patent protection.
From the above, it will be apparent that the process of obtaining patent protection for AI-based innovations involves challenges including the initial decision concerning whether to apply for protection, the preparation of the application documents and the prosecution of the patent applications in the countries of interest. These challenges are significantly more difficult than in many other areas of technology given the emerging nature of AI technology and the application of patent law to protecting AI technology.
A long and winding road
In conclusion, SMEs and growing start-ups should engage with patent protection from the start regardless of its complexities. The fact is, that every day a company waits to file a patent for a given invention is a day when someone else might beat it to the punch; something especially common when working with innovations in such popular fields as ML or AI.
Even if going ahead with patent applications is not possible from the get-go, it is valuable to make sure that any information that might, in the future, be relevant to a patent application is recorded with great accuracy and stored safely. Information that would tell a skilled person how to work the invention should be kept strictly confidential until after the patent application is filed, as it is generally impossible to get a patent for something if it is in the public domain.
A lack of industry-specific support is also no excuse to shy away from taking on the question of whether to patent AI algorithms head on. There is a growing number of professionals specialised in this algorithmic innovation and able to support SMEs, and as they go through the patenting process together with innovators, they are amassing valuable and practical knowledge that can benefit other companies too. Businesses need to seek this out when necessary, so they can turn patenting from a worrying headache to a crucial process in the growth and development of their venture.