Furthermore, these materials, interim deliverables, and deliverables have high values and therefore both the
vendor and the user have strong intentions to monopolize/re-use them.
For example, if we assume the typical case of the vendor who generates a trained model using raw data provided by the user, the intentions of the vendor and the user with respect to the vendor’s deliverables will conflict as shown below.
▼ User’s Position
The user wants to monopolize the training datasets and trained models since they are generated largely from the raw data that incorporates the user’s own know-how and trade secrets and since the user has paid a commission fee [to the vendor] at the time of development.
▼ Vendor’s Position
Since both the process of generating training datasets using raw data and the process of generating trained models using raw data use the vendor’s own advanced know-how and tremendous labor, and since the training data and the trained model are reusable, the vendor wants to use them for other development applications other than the contracted development project.
Therefore, a framework to reconcile these conflicting intentions is necessary.
So, how should this be handled?
With respect to this issue, I believe that the following type of framework can be used to reconcile [the conflicting intentions].
- (1) Among the materials, interim deliverables, and deliverables, know which are or are not covered by intellectual property rights
- (2) With respect to (1) above, know who has what rights under the default rules (i.e., the legal rules)
- (4) Know the limitations of the contract
▼ Let’s start with the general theory
First, before explaining this framework, I would like to explain the general theory about the relationship between law and contracts.
In seminars and other presentations, when I ask whether [the audience] understands the relationship between contracts and the law, most people understand that, in principle, when a promise is reduced to writing in a contract, even though it may differ from what is written in the law, that contract takes precedence [over the law].
On the other hand, there is an unexpectedly large number of people who are not aware that the provisions of the law automatically apply to matters not stipulated in a contract.
An illustration of this second point looks like this:
So, the diagrams above clearly show that when a matter is not stipulated in a contract, the provisions of law will automatically apply. Thus, if this “relationship between laws and contracts” is simply summarized in the matrix, the relationship looks as follows:
Let me provide a few concrete examples.
For example, let’s consider the case of an entrustment agreement which contains a contractual special provision for compensation of damages and the case where such contract does not contain such a provision.
Although the scope of damages for failure to perform an obligation is prescribed in Article 416 of the Civil Code, if there is a contractual provision that differs [from Article 416] (such as increasing the amount of damages), then such contractual provision will take precedence over the law.
Conversely, when there is no contractual provision relating to the scope of damages, [such matter] shall be in accordance with the provisions of law.
one more example.
Let’s consider the scope of a hardware vendor’s use of sensor data that has been generated by a device introduced into a user’s factory by the hardware vendor and sent to the hardware vendor with the user’s consent.
To start with, since sensor data is not covered by any intellectual property rights, such as copyright, there are no provisions under the current laws concerning who holds the right to use or who can use [such sensor data].
So, this falls under the “no legal provisions exist” area on the matrix above.
“Pristine state” essentially means that “the holder can freely use [the data]”; and in this example, it means that the vendor can freely use the data while the user cannot impose any restrictions on the vendor’s use of the data or have any right to demand the handover of the data.
▼ How about the case of AI development?
Then, if you apply this general rule to AI development, it becomes like this.
First is the point “(1) Among the materials, interim deliverables, and deliverables, know which are or are not covered by intellectual property rights.”
Next is the point “(2) With respect to (1) above, know who has what rights under the default rules (i.e., a legal rule).”
After grasping what items are covered by intellectual property right protection in point 1 above, you should understand the legal rules (the Patent Act, the Copyright Act, and others)relating to such covered matters in point 2.
The problem is how to design the legal default rules or the “pristine state” in the contract.
Although this will be discussed in detail later, this design method is the key point in the way of thinking about the rights and intellectual property in the AI development agreement and one of the most important points in the AI Guidelines.
Finally, there is the point “(4) Know the limitations of the contract”.