Image by tawatchai07 on Freepik

Japanese Ministry of Economy, Trade and Industry: Governance Guidelines for Implementation of AI Principles suggestively applied to the AI systems.  

The Japanese Ministry released version 1.1 on Jan 28, 2022 which is the work of an Expert Group on How AI Principles Should be Implemented. Like the U.S. AI Ethics Framework, these Guidelines are also a living, non-legally binding document with the aim of supporting the implementation of the AI principles, required to facilitate the AI deployment by providing the very comprehensive use case examples. 

1. Conditions and Risks Analysis:
   1. First, You should understand both potential positive and negative impacts of AI systems. The guidance underlines the fact that social acceptance of AI systems may be different in each country/region. For the systems applied to the great extent of consumers, the developers should take into account the social norms to assess if the outputs may cause incidents. 
   2. Second, same for the potential impacts, you should also understand the social acceptance of development and operation of the AI system itself. 
   3. Third, it is important to understand your company’s AI proficiency, which is your readiness for developing and operating AI systems. In other words, it is your ability to respond to their negative impacts.   
2. Goal setting – It may be helpful for a long run to consider setting AI governance goals based on the Human-Centric Social Principles of AI . 

3. System design – building an AI management system
   1. Firstly, it is an essential process to incorporate gap analysis between AI governance goals and current state of your AI systems in order to address the gap into the management system. In detail, you should (i) ensure your gap-analysis process is consistent with industry-standard and incorporate it into your own process; (ii) then, provide your users with sufficient information on potential gaps and measure to address them, including a contact point, (iii) finally, ensure that your data providers give your AI systems developers with sufficient information for gap analysis.      
   2. Secondly, it is crucial to improve literacy on AI ethics of human resources responsible for AI management systems, including for those in top management, employees engaged in data provision.
   3. Thirdly, according to the Principle of Fair Competition, you should encourage your developers, operators and data providers to cooperate to reinforce AI management by agreeing on measures to properly protect and disclose confidential information, including trade secrets. You may want to (i) understand the current state of information sharing between plural companies, (ii) encourage gathering information and exchanging views routinely for conditions and risks analysis.
   4. Finally, in this stage, you should reduce incident-related burdens on users by preventing incidents and through early response. Your company may (i) start by allocating burdens of addressing uncertainties amongst companies appropriately; and (ii) then consider in advance actions to take in response to incidents or disputes.        
4. Implement
   1. First of all, your AI systems developers and operators must be ready to explain about implementation status of AI management systems externally. They should record the gap analysis process and take other relevant actions. 
   2. Next, you may want them to ensure that they are also ready to explain the operating status of individual AI systems. 
   3. Finally, you should consider ranking implementation of AI governance as non-financial information pursuant to the Corporate Governance Code and proactively disclose such information.
5. Evaluation
   1. First, you should verify whether your AI management works appropriately to achieve the AI governance goals through the implementation of actions 3 and 4. 
   2. Then, you may want to seek the feedback from various outside stakeholders, including consumers, experts, NGOs, labour unions. 
6. Re-analysis of conditions and risks 

Finally, you should conduct re-evaluations and re-implementation of the actions 1(1) and (2) in a timely manner based on the opinion obtained in action 5(2).    

For us, check our previous post at the following: https://www.astraiagear.com/2023/01/17/ag-project-suggestions-by-the-authorities-around-the-world-for-ai-systems/

Sources: 

Guidances: https://www.meti.go.jp/shingikai/mono_info_service/ai_shakai_jisso/pdf/20220128_2.pdf

Expert report in 2021: https://www.meti.go.jp/shingikai/mono_info_service/ai_shakai_jisso/pdf/20210709_8.pdf