A Force for Good or Daring to Be Evil?

An Analysis of Google's Patent Policies on Open Source AI Software (3)

2018.11.29Daeho Lee

Google’s Open Patent Non-Assertion Pledge (OPN) is a pledge that explicitly declares which of Google’s patents are licensed to users as part of the Apache open source license, under certain conditions.

The OPN addresses the issue surrounding patents that we raised in the previous article of this series. Because open-source licenses typically refer to patents only in general terms, it is difficult for software users to determine which patents are covered by and which patents are excluded from the license.

Google's OPN exists as a supplement to the Apache license and explicitly lists those patents that are related to the relevant open source software and are included under the Apache license. However, the license for the patents is not unconditional, and the OPN includes provisions that permit Google to enforce these patents for defensive purposes. That is, if any user of Google’s open source software sues Google for patent infringement, Google reserves the right to cancel the license granted under the OPN and enforce its own patents against said user./p>

Nevertheless, the key point of the OPN that Google advertises is that the OPN allows users to use Google’s open source software more freely without having to worry about infringing Google’s patents.

Screenshot of the Open Patent Non-Assertion Pledge page

Google summarizes the benefits of the OPN as follows:

“Patent holders determine exactly which patents (and related technologies) they wish to Pledge and offer the public transparency in the process”
“Allows for defensive termination relative to a broader range of incoming patent attacks.”
“Non-assert promise and defensive use only terms designed to remain in force for the life of the patents, even if sold or transferred.”

The first benefit is that the OPN purportedly solves the problem that a user cannot easily know only by referring to the provisions of the license whether or not they are allowed to use a certain patent. Thus, Google is saying: “We hereby specify which patents we will not enforce (while reserving the right to enforce them under certain circumstances).” The intention is to reduce uncertainty among deep learning developers about the key deep learning patents owned by Google.

The second benefit is that the OPN still allows for enforcement of the pledged patents for defensive purposes. In other words, the patented technology is freely available to open source software users, but Google does reserve the right to enforce its patent rights for defensive purposes. That is, the patent will not be wielded as a weapon of attack, but if Google is attacked by a third party, Google will not pull its punches.

The third benefit is that even if these patents are sold to other companies, open source software users remain protected. These days, business transactions involving the buying and selling of patents are commonplace, and there are even companies that pursue a business model based on patent transactions. These so-called “patent trolls” buy patents from bankrupt companies or discontinued businesses and then use these to sue other companies for infringement, with the goal of profiting from a financial settlement. The OPN intends to obviate the risk faced by users of open source software of getting sued for infringing patents that they believed they were free to use, as a result of a sale of said patents.

The importance of this third benefit is felt strongly in Google’s ongoing dispute with Oracle. Oracle is a company that knows how to expertly and strategically use its intellectual property rights. In recent years, Oracle has been actively enforcing its intellectual property rights in Java, following its acquisition of Sun Microsystems. The OPN offers users protection.

Now that Google has decided to declare a list of patents for which it grants free licenses to open source users, we shall have to check whether or not the key patents related to deep learning are included in the list.

A list of patents declared by Google can be found at the following link:

However, to our surprise, this list includes not a single patent related to deep learning technology.

To avoid any misunderstanding, this is not to say that Google has expressed a willingness to enforce its rights existing in its deep learning patents.

Perhaps the list has not been updated yet, and an internal review to update the list may be in progress.

In any case, just because there are no deep learning patents on that list, it does not mean that Tensorflow users have to worry about all of Google's deep learning patents. That is because the Apache 2.0 license, which applies to the Tensorflow framework, already grants licenses for those patents essential to the use of the software, even without an explicit mentioning of the relevant patents in the OPN.

It would certainly be convenient and reassuring to users if the OPN included in its list the relevant deep learning patents, but at this point, we must still check Tensorflow's license regulations carefully and determine by ourselves whether or not a patent can be used freely.

In the final article of this series, we will return to one of the issues raised in our first article and analyze an example patent to determine whether it is free to use under the terms of the Apache 2.0 license.

A complete analysis of all relevant patents would unfortunately go beyond the scope of this series of articles, but we hope that the example discussed in our final article will provide some useful guidance.

About the author

Daeho Lee

Daeho started his career as a patent attorney at Seoul firm Park, Kim & Partner, where he worked on ...

Read more

Contributors for this article: