License: Creative Commons — Attribution-No Derivatives 4.0.
Open Source Software (OSS) is today an integral part of every software development project, and there are open source projects for almost every utility and functionality. Software relating to Artificial Intelligence (AI) is not an exception, and there are several AI linked open source initiatives targeting the automotive sector. From autonomous vehicles to telematics, there are many open source projects addressing different aspects of mobility. Some examples of specific AI projects for the automotive industry include Apollo Project, Udacity, StreetDrone and Autoware for autonomous automobiles; and Automotive Grade Linux, Eclipse Open Mobility, and Asset Performance-Prediction for shared mobility, electrification, infotainment, connectivity and so on. AI software like Tensorflow, CAFFE,Theano, Acumos and Skymind may also be used for general functionalities within all sectors, and are considered by the automotive sector as well for specific use cases. While the increase in the number of AI related open source for automobiles bodes well for the industry, it also brings forth challenges and risks with respect to intellectual property, data protection, safety, and security, among others. This note focuses on some important IP risks and challenges that are considered by companies while adopting and using Open Source Software with specific reference to the automotive industry.
Open Source Licenses
The license governing the software determines whether a software is open source or not and what IP flexibilities are available to a user. It is therefore one of the important starting points for addressing IP issues. There are about hundred open source licenses, which may be classified as viral, restrictive and flexible or permissive. Viral licenses like the General Public License (GPL)strictly enforce the copyleft philosophy, making it relatively difficult to implement proprietary models with respect to software developments, derivative works and integrations. On the other hand, permissive licenses like BSD, MIT and Apache provide significant latitude to use the software in business and proprietary contexts. They also permit companies and developers to protect and enforce intellectual property around the Open Source Software.
Most AI related open source projects have adopted existing open source licenses for the platforms, software, data, documentation and other subject matter created and made available by them. Continuing the growth in the use of permissive licenses after what could be called as the Android revolution, AI projects such as Apollo, Automotive Grade Linux and Autoware use Apache License or other permissive licenses. Some projects such as Udacity however continue to use GPL for their software. Considering the role of data in Machine Learning and AI, open data licenses such as Community Data License and Open Data Commons License have been created to share data. Having said that, some open data projects still use creative commons and MIT licenses for making data open and available.
Though the enforceability of open source licenses was ambiguous and most enforcement happened through community pressure, Courts have during the last decade upheld their enforceability. In a case involving a software relating to model rail road, a California Court not only upheld the validity of an open source license, the Artistic License, but also stated that non-compliance with the license terms gives rise to copyright infringement. The user of the Open Source Software in the case did not give appropriate attribution to the developer, and the said non-compliance gave rise to a cause of action for IP infringement rather than just breach of contract. Compliance of open source licenses while using the software therefore assumes high importance, and an easily avoidable mistake could prove to be very costly as was elucidated in the cases filed by Patrick McHardy, one of the initial contributors of Linux. Adopting an innovative strategy, McHardy sued several companies in Germany for non-compliance of GPL Version 2 with respect to use of Linux and effectively made money out of the cases and related settlements.
Software Modification and Integration
Compliance with license terms comes into play only when the Open Source Software is distributed or conveyed by the user. If Open Source Software is used for some functionality in a car, the sale of the car amounts to distribution and therefore, compliance becomes important. As per terms of most Open Source Software licenses, the modification of the Open Source Software or its integration with proprietary software increases the conditions to be satisfied. While viral licenses require modifications to be made available only under the specified open source license, permissive licenses allow users to distribute them under their own licenses. With respect to software under viral licenses, integrating proprietary code with open source code significantly restricts IP licensing flexibility.
Intellectual Property Ownership
As a general rule, there is no restriction against acquiring intellectual property protection over subject matter that forms part of Open Source Software, or is intended to form part of it. Open source licenses are in fact based on IP ownership and related rights. Inventors, developers and authors who release the initial software under an open source license, or subsequent contributors, who improve or build on the initial version can acquire patent as well as copyright protection over their software. Copyright and patent protection may also be acquired for aggregations, combinations and integrations of different software that are original, novel and inventive. For example, if X collects driver data on a large scale, Y creates a training data set, Z writes software for using the data set for predicting driver behaviour and makes it open source, and A, a car company, customizes and uses the software to improve driver safety in its cars, all parties in the chain can claim ownership over one IP or another. The car company can certainly consider patent and copyright protection for its specific implementation of Z’s software for driver safety in its cars. However, depending on the nature of the license governing Z’s software, what A can do with its contribution and the conditions it must follow may vary.
Having said that, in large open source projects involving several contributors, the question of who owns what may be difficult to ascertain. In one case filed by Patrick McHardy, the Court denied an injunction against Geniatech with respect to use of open source in satellite receivers and related license compliance because he could not prove contributions to Linux Kernel and copyright ownership relating to derivative works sought to be enforced. In the aforestated example, if Z is a community, it may not be easy to ascertain which part who contributed and thereby attribute IP ownership to different persons in the community to their specific contributions.
Patent assertion risks have always been part of Open Source Software usage. Though the initial developers and subsequent contributors grant patent licenses as part of licenses such as GPL and Apache, BSD style licenses are ambiguous with respect to patent license grants. Therefore, if a patent license grant is not clearly provided in the open source license, patent infringement risk from the developer/contributor subsists. Moreover, many open source licenses include very broad patent non-assertion clauses, which provide for termination of the license if patent enforcement action is taken by the licensee against the licensor. Such clauses pose enforcement problems for Open Source Software users, which could restrict the business and competitive value of their patent portfolios.
In addition to risks from parties to open source licenses, use of Open Source Software is also subject to patent risks from third party patent holders. Over the years, companies using Open Source Software were sued for, or threatened of patent infringement on several occasions. In an interesting case, Twin Peaks Software sued Red Hat for infringing its US patent with respect to the Mirror File System, and initially, Red Hat countered by arguing that the patent is not valid. However, later, Red Hat counter sued Twin Peaks for copyright infringement for using its version of Linux and not complying with GPL. In another case involving Versata, Ameriprise and XimpleWare, the case started with a claim for breach of contract, proceeded to non-compliance of GPL Version 2, and ended in a patent infringement suit, which was settled by Versata. The cases reflect some of the patent risks inherent in using Open Source Software and how complex they can get.
In 2005, an initiative called the Open Invention Network (OIN) was founded by IBM, Novell, Phillips, Red Hat and Sony with the objective of creating a pool of patents that will not be asserted against Open Source Software developers and users, and that will be used to defend patent infringement actions against the open source community. Its primary objective is to defend the Linux system and related software. As of date, OIN has more than two thousand members, which include several automotive companies. SpaceX, Honda, Daimler, Yamaha, Volvo and Ford are some of its recent members.
Pledging to not act against persons infringing patents for a particular purpose assumed significance in the automotive sector after Elon Musk’s blog post opening Tesla’s patent portfolio. In his 2014 post, Elon Musk stated that Tesla would not be filing patent law suits for using Tesla technology in good faith. Tesla later clarified that infringement actions against Tesla or electric vehicle companies, challenge to Tesla’s IP rights, and any copying of Tesla’s designs, products or IP would not be considered as good faith. In other words, a company using Tesla’s patents under the pledge has to effectively give up IP actions against Tesla.
While experts debate about the legal validity of Tesla’s patent pledge, Toyota recently opened up about twenty thousand of its patents under royalty free licenses to promote hybridisation of vehicles. At the same time, discussions to make open source implementations of standards are under process, but have not made much progress owing to their differential stands with respect to licensing standard essential patents. The developments however indicate the evolution of interesting and differential open source models in the automotive industry, and their significance for scientific/technological progress in the context of AI.
- Jacobsen v. Katzer, 535 F.3d 1373 (Fed. Cir. 2008)
- Ieva Giedrimaite, “Copyright Trolling: Abusive Litigation Based on a GPL Compliance”, The IPKat (February 24, 2019), available at http://ipkitten.blogspot.com/2019/02/copyright-trolling-abusive-litigation.html (last visited on June 18, 2019)
- Twin Peaks Software Inc. v. Red Hat, Inc., Case No. 5:12-cv-00911-RMW (N.D. Cal. Jan. 29, 2013).
- Versata, Ameriprise and XimpleWare cases –
- Versata Software Inc. f/k/a Trilogy Software, Inc. and Versata Development Group Inc. f/k/a Trilogy Development Group Inc. v. Ameriprise Financial Inc., Ameriprise Financial Services, Inc. and American Enterprise Investment Services, Inc., Case No. D-1-GN-12-003588; 53rd Judicial District Court of Travis County, Texas
- Versata Software Inc. v. Infosys, Case No. 1:10cv792, U.S. District Court, Western District of Texas
- Versata Software Inc. v. Ameriprise Financial Services Inc. et al., Case No. 1:14-cv-12, U.S. District Court, Western District of Texas, Case No. 1:14-cv-12, U.S. District Court, Western District of Texas
- XimpleWare Corp. v. Versata Software Inc., Trilogy Development Group, Inc., Ameriprise Financial, Inc., Ameriprise Financial Services, Inc., Aurea Software, Inc.,Case No. 3:13cv5160, U.S. District Court, Northern District of California
- XimpleWare Corp. v. Versata Software Inc., Aurea Software Inc., Trilogy Development Group, Inc., Ameriprise Financial Services, Inc., Ameriprise Financial, Inc., United HealthCare Services, Inc., Waddell & Reed, Inc., Aviva USA Corporation, Metropolitan Life Insurance Company, Pacific Life Insurance Company, The Prudential Insurance Company of America, Inc., Wellmark, Inc.,Case No. 5:13cv5161, U.S. District Court, Northern District of California (San Jose).
- Carlos Muñoz Ferrandis and Dr. Claudia Tapia, “Integrating Open Source into De Jure Standardization: Beyond a Call for the Appropriate License”, 53 les Nouvelles 221 (September, 2018).
- Lothar Determann and Bruce Perens, “Open Cars”, 32 Berkeley Tech. L.J. 915 (2017).
- David J. Kappos, “Open Source Software and Standards Development Organizations: Symbiotic Functions in the Innovation Equation”, 18 Colum. Sci. & Tech. L. Rev. 259 (Spring 2017).
- Stephanie Vu, “Pledging Patents Effectively: Copyright and Open Source As A Framework For Patent Pledges”, 14 Colo. Tech. L.J. 437 (2017).
- Benjamin M. Hill, “Powering Intellectual Property Sharing: How to Make Tesla’s Patent Pledge Effective”, 24 J. Intell. Prop. L. 191 (fall 2016).
- Matthew U. Scherer, “Regulating Artificial Intelligence Systems: Risks, Challenges, Competencies, and Strategies”, 29 Harv. J.L. & Tech. 353 (Spring 2016).
- John D Harkrider and Russell M Steinthal, “The Open Source Paradox: Innovation in the Absence of Exclusive Property Rights”, 7 No. 2 Competition L. Int’l 38 (November 2011).