Author: Masatoshi Tanaka (attorney at law/patent attorney)

１．What is happening about AI Data and Learned Models?

Nowadays, using AI and IoT (Internet of Things) to provide high value-added services and goods, to gain and maintain competitive advantage are indispensable perspectives in conducting business.

Although, the data used for AI and IoT and high quality learned model gained through AI learning have very high economic value, legislation such as on legal rights and legal protection procedure is still insufficient. We receive so many inquiries regarding this matter.

To protect these data and learned model, several tools are available by law.

Understanding “What kind of tools are there, and what are Pros and Cons of each tool” is important to maintain your business success.

In other words, if you pin down this matter and consider the business scheme in advance with this point of view, you can not only avoid many kinds of troubles, but also strengthen your business model by using AI and data.

Let’s look at the case below.

　　

This case shows that Development Company B is outsourced by Store A to develop a learned model which is able to f with high accuracy on what could be sold on when by machine learning together with POS cash register data of Store A and various social data such as date and time, weather, temperature, and whether the event is held or not. There is no problem providing learned model to Store A, but what will happen if Development Company B conduct the either of the following matters and Store A complains against Development Company B about it and the dispute occurred between Store A and Development Company B?

	①		Development Company B provides the competing store, Store C, with leaned model which is developed based on Store A’s POS data, and charges for usage fee.
	②		Development Company B is outsourced by Store C, which competes with Store A, to develop the similar learned model by using the learned model that is based on Store A’s data, and based on this, make the model to learn Store C’s POS cash register data. As a result, the more accurate learned model is developed at a lower cost in a short time.
	(3)		Development Company B creates data base of Store A’s POS cash register data, and provide Development Company D, which also creates similar system, with such data base.

２．Do AI’s learned model and data have copyrights?

Some clients say that “data” and “learned model” have copyrights. Unfortunately, we have to consider that basically there is no copyrights on “data” or “learned model.”

To be recognized as the (copyrightable) work, under the Copyright Act, the work has to be “a production in which thoughts or sentiments are expressed in a creative way”. In this case, on the other hand, “learned model” which nature is the function and “data” which just displays information are not recognized as “expressed in a creative way.”

We also have to note, however, that data organized according to certain rules and classification, such as learning data sets which are systematically constructed can be search using electronic computer, may be recognized its copyrightability as “database”.

However, “raw data” and “learned model” created by using and learning the learning data are not basically recognized their copyrightability. Therefore, they are not covered under the Copyright Act.

Thus, in the above case, although the data belongs to Store A, Store A only provided the raw data of its POS cash register, and therefore no one, including Store A, can possess copyright on this data.

３．Do AI’s Learned Model and Data have Patent?

In general, AI and learned model does not constitute Patent right because data and function itself are not applicable to the patentability of invention, “the highly advanced creation of technical ideas utilizing the law of nature.”

In case the data has a structure and can be processed by a certain electronic computer as a result of such structure, the patentability may be recognized. Also, in case a learned model is recognized as “program” by incorporating it into a specific process, and in case a learned model is deemed as “equivalent to the program” as a regulation of computer process, a patent may be established as a mean of information processor or its method of operation by combining with hardware. Please note that even in these cases, you have to apply for the patent before it goes to public, and not to disclose it to person who has no obligation of confidentiality prior to filing the application patent.

Thus, in the above case, because Store A only provided POS data, even there is a possibility that the patent is established, if Development Company B construct such data, Store A cannot obtain a patent right for the system including the learned model or make any claims against Development Company B unless it is stipulated in the contract or otherwise.

４．Can AI’s Learned Model and Data be protected as Trade Secret?

Under Unfair Competition Prevention Act, unauthorized disclosure, use and provision of trade secret is illegal.

To be protected as trade secret, the following three key factors are required:

1.　Kept secret/administered as secret (clear identification of intention to manage secrets and physical measures)
2.　Useful info (technical or business information useful for business activities)
3.　Not publicly known (info is only obtainable through the information owner)

In order to be recognized as “confidential”, it is necessary to b managed as secret/confidential from two perspectives below:

（1）Limited access to information (Access Limitation)
（2）The information is made known to those who have accessed the information that it is a trade secret (Recognizability)

If such strict confidentiality control is made, disclosure or use of such information without permission as a result of being protected as a “trade secret” may be regarded as a violation of the Unfair Competition Prevention Act and may be subject to injunction claims, compensation for damages or criminal penalties. In some cases, this could build a criminal case. As such strong effect is seen, the requirements such as administering as secret are also required strictly to some degree.

Thus, in the above case, if Store A strictly had kept secret POS data that they provide, Store A might be able to protect the database provided from Development Company B to Development Company D (act③ in the above picture).

However, even if POS data falls under “trade secret”, learned model itself, which is created from Store A’s POS data, is not included in Store A’s POS data. Therefore, diverting or repurposing such data (acts ①and ②in the above picture) we cannot say clearly that Store A is able to prevent using or repurposing such data directly based on the trade secret.

５．AI’s Learned Model, Data, and Contract

As described above, we could say that patents, copyrights, and trade secrets might partially protect AI’s learned model and data. However, this is still not entirely completed.

When the parties execute a contract regarding this matter, on the other hand, parties can make provisions as they like.

In the contract, we recommend you to carefully specify the handling of the learned model and data, additional development after the completion of development, pros and cons of diverting or repurposing, and data storage.

When drafting a contract using a general template, we have to say that the provision only mentioning “The intellectual property rights gained as a result of the development shall be shared.” is not enough. This is because, as we mentioned above, the intellectual property rights might not apply to AI’s learned model and data, and even if the patent rights apply thereto and to be shared, each joint owner can exercise such jointly owned patent. Thus, either joint owner cannot avoid the situation that the other joint owner diverts jointly owned patent rights without permission. 

When developing and sharing AI and data, establishing a contract scheme and specifying in a contract is important and necessary on how each “business model after development” and “data and learned model” can be used. 

“Always draft a contract carefully and properly regarding AI and Data beforehand.”

As a business owner, please keep in mind this and boost your business using AI and data.