Executive Briefing: The hidden data in AI IP

Three elements will control the AI market: technology, data, and IP rights. Leveraging rich patent data, Thomas Marlow uncovers the companies with the top patent holdings across the world in groundbreaking research and implementation technologies, surfacing insights into the sources and owners of AI technology as well as the hurdles and opportunities that those entering the field today face.

Over 500 years ago, the first patent statute was enacted in Venice, and patent systems evolved across the world from there over the next half-millennium. In the US, the constitution specifically authorizes a patent system through language found in Article 1, Section 8, Clause 8: “To promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries.” At the most basic, a patent represents a trade—a limited monopoly in exchange for promotion of science. What does this have to do with AI? Data. Every patent application must include a detailed technical description of the invention to be protected. This is in turn published publicly so others can read and learn from the inventor. Moreover, the patent also carries some very useful additional metadata: inventor names, addresses, the companies they work for (the patent owner), the date of the patent filing, a list of related patents/applications, and more. Between this metadata and the technical description of the invention, you have quite an amazing dataset identifying research and development activity that has been deemed valuable enough to spend money on for a chance at patent protection (not cheap). In many ways, if you want to analyze a technology space, patent data can be better than analyzing products; it provides much more detail and identifies technology that did not or has not hit the market. Drawing on an in-house dataset of over 100 million global patent matters and their metadata, Thomas Marlow digs into the tens of thousands of patents in force and pending before world patent offices, which demarcate the field of AI. This rich dataset includes ownership, inventorship, importance indicators, custom categorization into technology types and application fields, and additional categorization and analytical data is generated from a balanced combination of expert and AI-assisted analysis. Thomas dives into this data to illustrate technology development trends, who owns what, where this IP originated, where it is held, and how this data lines up with to the current rhetoric around AI superpower development. US and China AI happen to be a hot topic for discussion these days. Today China stands toe to toe with the US in its value from a patent holding and enforcement venue. This means that there are a lot of patent filings in China, providing great data to work off of when comparing US and Chinese development and the value of technology in each jurisdiction. While in the past, Chinese companies have repurposed Silicon Valley technologies for the China domestic market, the pendulum may swing back, based on data showing AI implementation technology patented by domestic Chinese companies in China but not the US. The publication of Chinese patents on technologies not concurrently protected in the US—which makes that technology public domain, able to be freely used—has the potential to drive US copying of Chinese-originated AI tech. Join in as Thomas pulls back the curtain, offering a glimpse at those technologies. Topics include: