(Credit: DrAfter123/iStockphoto.com)

In addition to issues regarding inventorship and infringement, companies utilizing artificial intelligence (AI) and machine learning may face obstacles when training data is obtained internationally. As mentioned in a previous post, many applications of machine learning continually train and update the relevant models and neural networks using data collected from a variety of sources. Often, products and services utilizing machine learning are sold or utilized internationally. Various issues can arise regarding the ability to protect inventions resulting from that international usage.

Public Disclosure

Outside the United States, many countries require that a patent application be filed before the invention is disclosed or used in public. If a product or service that relies on machine learning is used internationally, an invention generated by the AI may be produced or first used in a foreign country. If the invention is not discovered or anticipated before it is produced by the machine learning algorithm, and if it is produced or first utilized in a foreign country, there may be no ability to protect that invention outside the United States.

Thus, companies may need a process in place to anticipate or review any new processes before those processes are released or used outside the company. Otherwise, a company may need to limit new processes to initial usage in the United States, and then only roll out those inventions to other countries after review and protection. Some geolocation information may also be required for users who might take products or access services outside the U.S.

Foreign Filing License Requirements

As discussed in the previous post, data can be continually collected from users in order to train the models and networks. If some of this data is determined to significantly contribute to an invention produced by the machine learning algorithm, then the person submitting that data (or owning the device that submits that data) may be considered to be an inventor. In addition to the ownership concerns discussed previously, this can lead to problems with foreign filing licenses.

A user accessing a computer program while in India might generate a result that is analyzed by the AI and significantly contributes to a new invention. If the person is arguably an inventor based on that contribution, and the contribution was performed from India (and any other requirements met), a foreign filing license from India may be required before the invention can be protected in other countries. Companies may then need to limit the collection of data from certain international locations in order to avoid foreign filing issues. Further, if failure to obtain a foreign filing license can subject the inventor to criminal penalties, companies may be required to inform users of the risk in their terms of use.

Related Issues with International Data Collection

Companies may encounter additional risks to international data collection. For example, if a contribution from a user causes that user to be an inventor, then the user must be identified and listed as an inventor under current law. This can create a headache for companies, as the company must be able to determine not only which data or contribution led to the invention, but also the associated user or users. In addition to being potentially resource intensive, privacy policies may prevent tracking or disclosure of such information.

Similarly, users may have the option of either not having data collected or having the data collected “anonymously.” The question arises, then, as to whether this anonymous data can be used to train neural networks if the user will not be identifiable for inventorship purposes. There may be very complicated data privacy, export, and communications regulations and agreements between countries, which any company collecting this data would need to ensure are not violated. Companies may again need to track or determine the location of its users to determine whether any international data export would be involved, but it is possible that the obtaining of such location information itself may violate privacy or data collection laws.

While the ability to collect data and information from users and actual customer usage can greatly improve the accuracy of machine learning and produce new and interesting technology, that ability comes with various risks to the companies providing the algorithms, as well as the users themselves. There can also be difficulty in protecting that technology. Until the laws are updated to address these situations, companies should be mindful of the data that is collected and used for machine learning, and the potential implications.

In addition to issues regarding inventorship and infringement, companies utilizing artificial intelligence (AI) and machine learning may face obstacles when training data is obtained internationally. As mentioned in a previous post, many applications of machine learning continually train and update the relevant models and neural networks using data collected from a variety of sources. Often, products and services utilizing machine learning are sold or utilized internationally. Various issues can arise regarding the ability to protect inventions resulting from that international usage.

Public Disclosure

Outside the United States, many countries require that a patent application be filed before the invention is disclosed or used in public. If a product or service that relies on machine learning is used internationally, an invention generated by the AI may be produced or first used in a foreign country. If the invention is not discovered or anticipated before it is produced by the machine learning algorithm, and if it is produced or first utilized in a foreign country, there may be no ability to protect that invention outside the United States.

Thus, companies may need a process in place to anticipate or review any new processes before those processes are released or used outside the company. Otherwise, a company may need to limit new processes to initial usage in the United States, and then only roll out those inventions to other countries after review and protection. Some geolocation information may also be required for users who might take products or access services outside the U.S.

Foreign Filing License Requirements

As discussed in the previous post, data can be continually collected from users in order to train the models and networks. If some of this data is determined to significantly contribute to an invention produced by the machine learning algorithm, then the person submitting that data (or owning the device that submits that data) may be considered to be an inventor. In addition to the ownership concerns discussed previously, this can lead to problems with foreign filing licenses.

A user accessing a computer program while in India might generate a result that is analyzed by the AI and significantly contributes to a new invention. If the person is arguably an inventor based on that contribution, and the contribution was performed from India (and any other requirements met), a foreign filing license from India may be required before the invention can be protected in other countries. Companies may then need to limit the collection of data from certain international locations in order to avoid foreign filing issues. Further, if failure to obtain a foreign filing license can subject the inventor to criminal penalties, companies may be required to inform users of the risk in their terms of use.

Related Issues with International Data Collection

Companies may encounter additional risks to international data collection. For example, if a contribution from a user causes that user to be an inventor, then the user must be identified and listed as an inventor under current law. This can create a headache for companies, as the company must be able to determine not only which data or contribution led to the invention, but also the associated user or users. In addition to being potentially resource intensive, privacy policies may prevent tracking or disclosure of such information.

Similarly, users may have the option of either not having data collected or having the data collected “anonymously.” The question arises, then, as to whether this anonymous data can be used to train neural networks if the user will not be identifiable for inventorship purposes. There may be very complicated data privacy, export, and communications regulations and agreements between countries, which any company collecting this data would need to ensure are not violated. Companies may again need to track or determine the location of its users to determine whether any international data export would be involved, but it is possible that the obtaining of such location information itself may violate privacy or data collection laws.

While the ability to collect data and information from users and actual customer usage can greatly improve the accuracy of machine learning and produce new and interesting technology, that ability comes with various risks to the companies providing the algorithms, as well as the users themselves. There can also be difficulty in protecting that technology. Until the laws are updated to address these situations, companies should be mindful of the data that is collected and used for machine learning, and the potential implications.