Japan’s Instructions For Generative AI Services

Generative AI continues to expand in both capabilities and possibilities. Whether it’s writing a modern Shakespearean play or producing a unique source code for an interactive website, it can all be done within a few seconds.

While this phenomenon presents many opportunities, one, particularly critical challenge it poses is its reliance on a nearly endless stream of data collection. The more data it collects and learns, the more efficient and effective it becomes.

However, data collection on such a scale is a highly volatile proposition with an increased likelihood of users’ data privacy rights being violated. Governments worldwide have become increasingly active in regulating various aspects of AI, particularly its data collection mechanisms.

Brief Background

In Japan, on May 20, 2023, the Personal Information Protection Commission (PPC) laid down important considerations and instructions for all major stakeholders in relation to Generative AI services (services that use AI to generate sentences, images, etc., in response to questions, work instructions (prompt input), etc.).

These include instructions for business operators, administrative agencies, and regular users that use such generative AI services. These instructions are meant to secure the appropriate rights and interests of users through the appropriate handling of users’ personal data in line with the Act on the Protection of Personal Information (APPI) within the context of generative AI services.

Based on Article 147 of the same Act, a blanket warning was issued to all generative AI services, particularly OpenAI that developed ChatGPT, regarding their collection, storage, and usage of users’ personal data and information.

Lastly, the Committee also states that as new concerns are recognized in the future, additional measures and instructions will be released as necessary, owing to the highly dynamic nature of generative AI services and their data collection capabilities.

Instructions for Business Operators Handling Personal Data

Here are the instructions specific to the use of generative AI services by business operators:

• If personal data or information needs to be provided to a generative AI service, the business operator must ensure that the content of such personal data is limited to that necessary to achieve a specified purpose;
• No personal data may be provided to a generative AI service for usage in a context not consented to by the user, as this may result in a violation of the Act on the Protection of Personal Information (APPI);
• Similarly, the usage of personal data for any purposes other than generating responses to a prompt, such as machine learning, may also lead to the violation of the APPI.

Instructions for Administrative Agencies Handling Personal Data

Here are the instructions specific to the use of generative AI services by administrative agencies:

• If personal data or information needs to be provided to a generative AI service, the administrative agency must ensure that it provides as minimal personal data as possible for the specified purpose;
• As required for businesses, administrative agencies should also ensure that any personal data provided to a generative AI service should not be used for any purpose other than generating responses to prompts;
• The administrative agencies must ensure that any organization providing such generative AI services does not retain any personal data or information for machine learning purposes.

Instructions for General Users

Here are some key considerations related to the use of generative AI services by the general population as released by the Committee:

• The generative AI service may use the personal information provided for its own machine-learning purposes. Moreover, such information may also be statistically linked with other information and output from the service with no guarantee of accuracy of the content. Therefore, users should be cognizant of the risks involved when providing personal information to generative AI services;
• The generative AI services may produce responses that contain inaccurate content, as the sentences generated by such services are based on probabilistic correlation. Thus users should make appropriate judgments based on such risks;
• Additionally, the users must fully go through and check the terms of use, privacy policy, and other notices/disclosures that detail the functions of the generative AI services to make appropriate decisions regarding what information should be provided to the service.

Additional Requirements

Here are some additional requirements placed on generative AI services by the Committee:

• If personal data or information needs to be collected for machine learning purposes, the following four steps must be thoroughly undertaken:
  • Ensure that personal information requiring special care is exempt from the information being collected;
  • As soon as possible after the information is collected, take measures to ensure that as much quantum of personal information requiring special care is reduced as possible;
  • If, even after the implementation of the foregoing measures, it is discovered that the collected information still contains personal information requiring special care, initiate prompt actions to ensure such information is deleted or it is made impossible to identify a specific individual through such information before it is used as part of a training dataset;
  • If a user or the Personal Information Protection Commission (PPC) requests or instructs a halt to the collection of personal information from a specific site or third party, such requests or instructions must be complied with unless there is a justifiable reason for refusal.
• A user should be allowed to consent to the use of their personal information requiring special care for machine learning. If a user does not consent to the foregoing processing purpose, the generative AI service should not use their personal information for this purpose unless there is a justifiable reason.

How Can Securiti Help

Generative AI services rely on data collection. It’s the life and blood of this technological leap. However, it is just as important for these services to adhere to their data collection and processing obligations in light of the plethora of data protection legislation globally.

These laws place different obligations on those collecting and processing the data. As specific AI laws and regulations are only expected to proliferate moving forward, automation offers the best way for generative AI services to remain compliant with their data protection obligations.

Securiti is a leader in providing enterprise data privacy, security, governance, and compliance solutions.

For example, the PrivacyCenter.Cloud supports enterprises in their journey toward compliance with the Japanese APPI through automation, enhanced data visibility, and identity linking. Not only does it allow for efficient compliance, but it does so in an incredibly user-friendly manner, with its central dashboard being both easy to navigate and use.

Request a demo today and learn more about how Securiti can help you comply with the APPI and any other major data regulation globally.