Cloud Hosting

Artificial intelligence (AI) has quickly become one of the pillars of the modern economy. According to one widely cited study from 2017, AI could contribute up to $15.7 trillion dollars to the global economy by 2030. See PwCs Global Artificial Intelligence Study: Exploiting the AI Revolution at 3, https://www.pwc.com/gx/en/issues/data-and-analytics/publications/artificial-intelligence-study.html. That prediction is already coming to fruition. According to a White House report on AI from February 2020, AI is already having a substantial economic impact, not only for companies whose core business is AI, but also for nearly all other companies as they discover the need to adopt AI technologies to stay globally competitive. American Artificial Intelligence Initiative: Year One Annual Report (Feb. 2020) at 1, https://www.whitehouse.gov/wp-content/uploads/2020/02/American-AI-Initiative-One-Year-Annual-Report.pdf. The recognition of the importance of AI is both broad and worldwide. Russias Vladimir Putin has gone as far as to state that whoever becomes the leader in [AI] will become the ruler of the world. See The Verge (Sept. 4, 2017), https://www.theverge.com/2017/9/4/16251226/russia-ai-putin-rule-the-world.

It is thus no surprise that companies are heavily investing to protect the intellectual property generated from their investments in AI technology. See PwC MoneyTree Report (Q4 2018), https://www.pwc.com/us/en/moneytree-report/moneytree-report-q4-2018.pdf. The question becomes how best to protect those investments in this critical space. For example, an autonomous driving company may be looking at its AI training data (i.e., records of previous test drives), the artificial neural network implementations generated from that training data (i.e., the software that helps the car drive itself), and assortments of other data necessary to operate an autonomous car. For each of these elements, the company must examine what aspects are patentable, subject to trade secret protectionor both. A misstep could result in the company being left with no meaningful intellectual property protection for its most important research and development.

But patenting AI technology today can be difficult. Due to the prohibition on patenting abstract ideas, acquiring meaningful patents on artificial intelligence systems is not straightforward. Thus, companies are increasingly turning to trade secret protection to protect their AI-related intellectual property. This article explores the tradeoffs between patents and trade secrets in the AI sector. It then describes how trade secrets have become essential tools for companies to protect their AI-related intellectual property. Finally, it concludes with practical guidance on how to leverage both patents and trade secrets to best protect valuable intellectual property regarding AI.

What is Artificial Intelligence?

First, a word on terminology. Some companies invoke the term artificial intelligence to describe their products at the earliest opportunityeven when the underlying technology does not fit within the established definition of artificial intelligence. For the purpose of this article, artificial intelligence generally refers to technology that, in some sense, mimics human intelligence. In particular, AI under this definition permits computers to perform some task without being expressly programmed to do so. To that end, this article will focus on machine learning, neural networks, and related training models, algorithms and data.

Patents Versus Trade Secrets The Tradeoff of Public Disclosure

Patents confer a legal right to exclude others from making, using, selling, and importing into the United States the claimed invention for a number of years. But, in order to take advantage of this government-sanctioned monopoly, the inventor must disclose the invention to the public with enough detail such that the invention can be recreated by others in that field. This quid-pro-quoa disclosure of the invention to the public in return for a limited-in-time monopoly on the inventionis a fundamental underlying policy objective of U.S. patent law.

By contrast, trade secrets, as the name suggests, protect information that is secret. Trade secrets can provide protection for any information where the owner has taken reasonable efforts to keep such information secret and the information derives independent economic value, actual or potential, from not being generally known to other persons. See, e.g., 18 U.S.C. 1839(3) (Federal Defend Trade Secrets Act, definition of trade secret); Cal. Civil Code 3426.1(d) (California Uniform Trade Secrets Act, definition of trade secret). Both federal and state law provide protection for trade secrets. Historically, trade secret protection has been applied to a wide variety of subject matter, including compilations of public data, source code, schematics, diagrams, and customer listsamongst many other pieces of information.

In many ways, trade secret law can be broader or more flexible than patent law. Unlike patents, trade secret protection can be obtained without any application or registrationit arises automatically if the trade secret owner takes appropriate steps to ensure the information is secret, so long as the information provides a competitive benefit. Trade secret protection can also theoretically last as long as the information is kept secret. And trade secret law protects items which would not be proper subjects for consideration for patent protection under 35 U. S. C. 101. Kewanee Oil Co. v. Bicron Corp., 416 U.S. 470, 482-3 (1974). For example, a list of customers could be protected as a trade secret, but certainly not by a patent.

In other ways though, trade secret protection is weaker than patent protection. Importantly, independent development is a defense to trade secret misappropriation but not for patent infringement. As explained by the Supreme Court in 1974:

Trade secret law provides far weaker protection in many respects than the patent law. While trade secret law does not forbid the discovery of the trade secret by fair and honest means, e. g., independent creation or reverse engineering, patent law operates against the world, forbidding any use of the invention for whatever purpose for a significant length of time. The holder of a trade secret also takes a substantial risk that the secret will be passed on to his competitors, by theft or by breach of a confidential relationship, in a manner not easily susceptible of discovery or proof. Where patent law acts as a barrier, trade secret law functions relatively as a sieve.

Kewanee, 416 U.S. at 489-490 (footnote and citation omitted).

This view is not universal. One can ask whether Coca-Cola, the holder of one of the most famous trade secretsthe formula for Coca-Colawould agree with it. More to the point, times have changed since the Kewanee decision in the 1970s. For certain types of innovations related to AI, the pendulum may be swinging away from patent protection and towards trade secret protection.

The Difficulties in Patenting AI Alice and Abstract Ideas

Recent years have seen a rapid acceleration in the number of patent applications directed to inventions in the field of artificial intelligence. More than half of all AI-related patent applications have been published since 2013. See WIPO Technology Trends 2019, Artificial Intelligence, at 13 https://www.wipo.int/edocs/pubdocs/en/wipo_pub_1055.pdf. Within that time, applications related to machine learning have grown by an average of 28% each year, applications related to computer vision have grown by an average of 46% each year, and applications related to robotics and control methods have grown by an average of 55% each year.

Despite this surge in applications, however, there are potential pitfalls to seeking patent protection over AI-related inventions. In particular, to receive a patent, the patent must claim patent-eligible subject matter under 35 U.S.C. 101. One category that is ineligible for patent protection is abstract ideas. Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014). Over the last 15 years, the general trend in the case law has been to apply the prohibition on patenting abstract ideas more strictly to software-centric inventions.

Given the limitations articulated in Alice and its progeny, it is unclear how many of the AI-related patents that have made their way through the U.S. Patent Office would survive in eventual litigation. See, e.g., Hyper Search, LLC v. Facebook, Inc., No. CV 17-1387-CFC-SRF, 2018 WL 6617143, at *10 (D. Del. Dec. 17, 2018) (invalidating patent with neural network module); see also Purepredictive, Inc. v. H20.AI, Inc., No. 17-CV-03049-WHO, 2017 WL 3721480, at *5 (N.D. Cal. Aug. 29, 2017),affd sub nom.Purepredictive, Inc. v. H2O.ai, Inc., 741 F. Appx 802 (Fed. Cir. 2018) (invalidating patent directed to automating predictive analytics). While each of these patents stands on its own and the decisions do not indicate that any future AI-related patents are necessarily invalid (or valid), they stand as guideposts that companies should be mindful of when considering patenting artificial intelligence inventions.

Finally, some AI-related innovations are simply not eligible to receive patent protection at all. For example, raw data collected for use in machine learning algorithms is not patentable in and of itself. That raw data combined with a conventional and well-known machine learning algorithm may also be unpatentable, even though the result may be incredibly valuable to the company. Considering these risks, many companies are turning to alternatives to protect their intellectual property in the AI spacenamely, trade secrets.

Trade Secrets An Apt Tool for Protection of AI Intellectual Property

While it is impossible to know the number of AI trade secrets being closely held by organizations around the world, it is likely that most AI intellectual property generated in the United States today is being protected through the use of trade secrets. While specific details remain confidential in light of strict protective orders, courts have already indicated that certain areas of information related to AI are protectable as trade secrets, such as algorithms, source code, and the way a business utilizes AI to implement machine learning.

There are certain practical advantages to trade secret protectionno filings fees, protection in real-time, theoretically unlimited length of protection, and broadly eligible subject matter. For AI in particular, there are several reasons why trade secrets are particularly valuable and suitable for intellectual property protection as compared to patents:

While trade secrets are increasingly important for AI companies, one major drawback in utilizing trade secrets is that protection is only afforded to the extent the intellectual property can be kept secret. Keeping software a secret can be challenging and operationally taxing for several reasons: (1) given the turnover at technology companies, strong employment agreements are needed to ensure departing employees are legally required to keep trade secrets secret; (2) given the ease of stealing softwarewhich can be as easy as downloading code to a USB drivestrong cybersecurity policies need to be created and enforced; (3) because reverse engineering can be a defense to trade secret misappropriation, software needs to be designed and deployed in a way to ensure reverse engineering is not possible (see, e.g., Sargent Fletcher, Inc. v. Able Corp., 110 Cal. App. 4th 1658, 1670 (2003) (Evidence of independent derivation or reverse engineering directly refutes the element of use through improper means.); N. Am. Deer Registry, Inc. v. DNA Sols., Inc., 2017 WL 2402579, at *7 (E.D. Tex. June 2, 2017) (trade secret is not misappropriated if there is reverse engineering or independent derivation)); and (4)in order to conduct business, it is often necessary to share technology widely with employees and partners, which increases the risk that a trade secret could be disclosed publicly.

In light of these concerns, maintaining trade secret protection can incur meaningful costs for a company and requires significant ongoing vigilance. Ultimately, a trade secret is only protected so long as it remains a secret. Even with strict regulations in place, companies always run the risk that the information will become public.

Patent vs. Trade Secret: Making the Right Decision for AI-Related Inventions

Even though trade secrets are important to protect AI-related intellectual property, there remain different advantages and drawbacks for both patents and trade secrets. The decision whether to patent or keep as a trade secret a given innovation thus represents an important strategic decision for any company. Here are some guiding factors to consider when making these kinds of critical decisions:

Is the innovation eligible for patent protection? Does the innovation satisfy the requirements of the Patent Act, including being patent-eligible subject matter under 35 U.S.C. 101? If not, then patents are unavailable and trade secret protection is the best option.

Does the innovation comprise the type of information that can be kept secret as part of your business? If the innovation is readily discernable from the product itself or by other appropriate means, trade secret protection is unavailable. The trade secret in that instance would not be secret. Thus, patent protection would be the best option.

Is the innovation likely to become generally known soon? Trade secrets only protect information that is not generally known. If the innovation is one that competitors or academia is likely to be making public relatively soon, then trade secret protection is sub-optimal. Instead patent protection may be the best option.

How likely is the patent able to withstand an attack in litigation? Even if the patent may be approved by the patent office, if you believe the patent is unlikely to withstand an attack in litigation, it may be better to keep the innovation as a trade secret so the underlying intellectual property does not have to be disclosed to the public.

How quickly will the invention become obsolete? If the invention will become obsolete quickly, the length of protection that patents provide (and the cost and effort to file the patent), may not be worth the benefit.

How quickly can the invention be commercialized? Conversely, if the invention will take a long period of time to monetize, the length of protection afforded by a patent will allow time for long-term investment and capitalization.

Is the innovation worth patenting? Patents cost time and money to prosecute and obtain. Not all innovations are worth that effort. For certain types of know-how, it may be more practical to utilize trade secrets to protect the innovation rather than a patent.

Both patents and trade secrets offer powerful ways for companies to protect their AI-related intellectual property. Each can be effective in certain circumstances. In most cases, optimal protection strategies will involve a thoughtful use of both regimes.

Read more:
April 2020: The Increasing Importance of Trade Secret Protection for Artificial Intelligence - JD Supra