On January 15, 2021, Japan’s Ministry of Economy, Trade, & Industry (METI) issued an interim report, AI Governance in Japan (English version here). The report, which was written by an expert committee, reviews global AI governance trends and considers options for turning “principles into practice” in the Japanese market. The USJBC and the Chamber’s Project Data-Driven Innovation, submitted formal comments on this report, and you can find them here (English), and the Japanese translation here.
The U.S. Chamber of Commerce and the U.S. -Japan Business Council welcome the opportunity to comment on the Ministry of Economy, Trade, & Industry’s Interim Report on Artificial Intelligence Governance in Japan.
The U.S. Chamber of Commerce (“Chamber”) is the world’s largest business federation, representing the interests of more than three million enterprises of all sizes and sectors. Through our U.S. -Japan Business Council (“USJBC”), we represent a substantial share of U.S. economic activity with Japan and advocate for stronger commercial ties between our two countries. The Chamber is also a leading business voice on digital economy policy, including on issues of artificial intelligence (“AI”), data privacy, cybersecurity, digital trade, and e-commerce. In the U.S., Japan, and globally, we advocate for sound policy frameworks that support economic growth, promote consumer protection, and foster innovation.
The Chamber and USJBC believe in AI’s potential as a force for good to tackle challenges such as the COVID -19 pandemic and to spur economic growth for the benefit of Japanese consumers, businesses, and society. In 2019, we issued ten principles for policymakers considering action on artificial intelligence:
1. Recognize Trustworthy AI is a Partnership
2. Be Mindful of Existing Rules and Regulations
3. Adopt Risk -Based Approaches to AI Governance
4. Support Private and Public Investment in AI Research an d Development
5. Build an AI -Ready Workforce
6. Promote Open and Accessible Government Data
7. Pursue Robust and Flexible Privacy Regimes
8. Advance Intellectual Property Frameworks that Protect and Promote Innovation
9. Commit to Cross -Border Data Flows
10. Abide by International Standards
We hope these principles serve as a guidepost for the Ministry of Economy, Trade, & Industry (“METI”), the Expert Group on Architecture for AI Principles to be Practiced (“Expert Group”), and the broader Government of Japan (“GOJ”). In addition to these general principles, the Chamber and USJBC have prepared specific input on the Interim Report on AI Governance in Japan (“Report”), below.
I. AI Governance & the U.S. -Japan Digital Partnership
The U.S. -Japan partnership is the cornerstone of a global digital economy based on open digital architectures, high -standard trade rules, and cross -border data flows. The U.S. and Japan have leverage d this partnership to advance common digital policy priorities in multilateral forums as diverse as the G7, the G20, the Asia -Pacific Economic Cooperation, and the World Trade Organization. At the heart of this partnership are strong bilateral ties built on a commitment to regular dialogue on digital policy and on frameworks, such as the U.S. -Japan Digital Trade Agreement. Appropriate focus on the bilateral digital partnership strengthens our commercial ties, while creating high standards that can be replicated elsewhere.
With President Biden’s Administration now in place, the U.S. Government and the GOJ ha ve an opportunity to consider the next chapter in our bilateral digital partnership. In addition to the traditional focus on digital trade, data flows, privacy, and cybersecurity, U.S. -Japanese cooperation on careful, pragmatic, innovation friendly, and risk -based approaches to AI governance should be a priority. Importantly, these efforts can be built on a strong multilateral foundation, as U.S. and Japanese collaboration was crucial to finalizing the Organization for Economic Cooperation & Development’s (“OECD”) AI Recommendation s and in establishing the Global Partnership on AI. Nonetheless, the Chamber and USJBC recommend that the GOJ and the Group of Experts explicitly identify areas for bilateral U.S. -Japan cooperation on AI governance. One opportunity may be in the development of voluntary governance frameworks. We note that the U.S. National Institute for Standards & Technology (“NIST”) has developed principles on explainable AI and has been directed by the U.S. Congress to develop an AI risk management framework, complementing its existing works on cybersecurity and privacy. Another area for cooperation may lie in establishing new dialogues or retrofitting existing ones between regulators, including in financial services, transportation, healthcare, and data protection, to discuss AI governance as applied to specific sectors and regulatory considerations. The U.S. Government’s implementation of its Guidance for Regulation of AI Applications may serve as an important reference point.
The Chamber and USJBC are committed to supporting U.S. -Japan cooperation on this essential matter of digital policy and welcome follow up discussions with the Group of Experts and the Government of Japan. Last year, we launched the Forum on Global AI Governance, a project undertaken jointly with Keidanren and Orgalim to bring together business leaders and policymakers from the U.S., Japan, and the European Union to discuss interoperable approach es to AI governance. We note that representatives from METI and the Ministry of Internal Affairs & Communications (“MIC”) joined these discussions. As we continue this project in 2021, the Chamber and USJBC seek to further support the GOJ’s leadership on issues of AI governance.
II. Multi -Stakeholder & Evidence -Based Approaches
The Chamber and USJBC recognize that fostering public trust and trustworthiness in AI technologies is necessary to advance their responsible development, deployment, and use. Given the speed and complexity of technological change, however, governments cannot foster trust in AI alone; rather, a partnership between governments, the business community, academia, and other stakeholders is needed. The Group of Experts should explicitly endorse transparent, multi -stakeholder approaches to AI governance, including in the development of voluntary standards, frameworks, and codes of practices to bridge AI principles and practice. We note that multi-stakeholder initiatives have the greatest capacity to identify gaps in AI outcomes and capacities and to mobilize AI actors to address them. One avenue for multi-stakeholder efforts is developing metrics by which organizations can measure an AI application’s performance according to agreed -upon metrics such as privacy. As acknowledged in the Report, the OECD maintains an AI Policy Observatory that has catalogued such initiatives around the world; it should serve as a global reference point for all future AI governance initiatives.
As the GOJ and the Group of Experts consider future action on AI governance, we encourage it to highlight the importance of robust evidence on which to base policymaking. We recommend that the Report highlight the role that principles of scientific integrity, information quality, and technical analysis must play in AI governance discussions. These principles are necessary when considering issues of public concern related to AI, as a rigorous evidence base is needed to advance AI governance in a targeted, balanced, and well -informed manner. They are also included in the U.S. Guidance for Regulation of AI Applications and may serve as an area for further alignment between the United States and Japan. Indeed, AI governance frameworks should address concrete harm to individuals that can be empirically linked to the use of AI technologies. Without an appropriate evidence base, future AI proposals may unintentionally stymie Japan’s digital transformation.
III. Flexible, Proportionate, & Risk -Based Governance Frameworks The Chamber and USJBC agree with the Group of Experts that AI governance frameworks should be risk -based, rather than prescriptive, and that the GOJ should avoid issuing binding one -size -fits -all solutions, which may inhibit the development and deployment of AI applications. AI governance frameworks must be flexible. The same AI application, put to the same use, will pose different risks depending on the way it is integrated into a business ’s operations, as the degree of human oversight and safeguards, such as monitoring, may vary between contexts and may increase or decrease the risk based on the AI in question. In this respect, the size of a company may be a poor measure of risk, as a small technology startup processing highly sensitive data may present higher risks than an established company using an AI at scale. Consequently, a sector or industry profile may n ot be an absolute measure of inherent risks, while a recognition for companies’ different roles as developers and deployers of AI applications may be needed.
The Chamber and USJBC therefore support the approach taken by the Group of Experts in favor of non -binding intermediary guidelines. Non -binding guidelines, developed with input from all stakeholders, are more likely to be adopted by businesses, given that there are few specific standards against which AI solutions may be measured at present. In this respect, the Report points out that these non -binding guidelines may be used as a minimum standard in certain regulatory monitoring and supervision. If this is the case, the GOJ should ensure that all supervising government agencies share the same understanding and expectations when creating and implementing the non -binding standards. As the Report’s “future issues” section suggests, monitoring and enforcement should be carefully designed and executed to balance the need to promote innovation, protect consumers, and minimize companies’ compliance burden.
The Group of Experts may also wish to further clarify specific risks that may be of concern (e.g., unintended unfair treatment of customers, lack of transparency, unclear lines of accountability). We recommend that the Report explicitly cite the importance of considering tradeoffs in a risk analysis, such as the opportunity cost of not adopting AI and whether the benefit of using AI in a specific context outweighs its potential har m. A discussion of AI -related risks should be sufficiently high -level and recognize the practical considerations of AI governance, as businesses are often the best positioned to weigh tradeoffs around implementation.
IV. An Enabling AI Policy Environment As a dynamic and rapidly evolving suite of technologies, AI requires a predictable and enabling policy environment to thrive. We note that AI does not exist in a legal vacuum, and that activities performed and decisions aided by AI are often already accountable under existing laws. As the Group of Experts correctly point out, there are multiple layers of AI governance, from multilateral agreements to hard and soft laws to technical standards. Consequently, AI governance frameworks must be simple and should avoid adding unneeded complexity to an ever -crowded regulatory landscape.
The GOJ should therefore appropriately study and account for existing rules — whether binding or not — before proposing new initiatives. Moreover, existing laws, regulations, regulatory authorities, and frameworks should be reviewed to ensure that, with respect to AI governance, they are coherent, consistent, and streamlined. On issues of data protection, even non -binding guidance should be coherent with the Personal Information Protection Commission’s interpretation and enforcement of the Act on the Protection of Personal Information. Failure to appropriately account for these rules before instituting a new governance framework or requirements may lead to overlapping and contradictory obligations in areas as diverse as financial services, healthcare, transportation, and data protection. At the same time, METI and the Group of Experts should encourage regulators to embrace experimentation in AI applications while also still ensuring consumer protection. Regulatory sandboxes and constructive dialogue with the business community would play a necessary role in this respect, and the lessons and insights that result from them should be accounted for when developing AI governance frameworks.
The Chamber and USJBC thank METI for the opportunity to provide these comments. The American private sector is ready to serve as an open partner to the Group of Experts and the GOJ as it considers the future of Japan’s AI governance framework. We are committed to assisting you in your efforts and hope our recommendations will be given due consideration.
U.S. -Japan Business Council
U.S. Chamber of Commerce
Senior Vice President
International Regulatory Affairs
U.S. Chamber of Commerce