UN Under-Secretary-General Tshilidzi Marwala States in T4IS2026 Keynote: 'AI Can Support 79% of SDG Targets, but Only 18% Are on Track'—Presents Four Targets for AI Regulation

Event Overview

Socias Co., Ltd. (Headquarters: Chuo-ku, Tokyo; CEO: Sera Yoon) hosted the keynote speech 'Diplomacy and AI: The UN Perspective' by Mr. Tshilidzi Marwala (Rector, United Nations University / UN Under-Secretary-General) on the main stage of the invitation-only executive summit 'Tech for Impact Summit 2026' (T4IS2026), held on Sunday, April 26, 2026, at the Kioi Conference, Tokyo Garden Terrace Kioicho.

Past summits have featured speakers such as Ms. Audrey Tang (former Digital Minister of Taiwan), Mr. Hirai (first Digital Minister of Japan), Mr. Charles Hoskinson (Founder of Cardano), and Ms. Kathy Matsui (Co-founder of MPower Partners, former Vice Chair of Goldman Sachs Japan).

As the EU AI Act enters a phased, full-scale application phase from August 2026 and AI governance policies are significantly advancing in Japan, Mr. Marwala delivered a 15-minute intensive lecture on the framework of global AI governance from the United Nations University based in Tokyo.

Core Messages of Mr. Marwala's Lecture

1. The '79% vs. 18%' Gap between SDGs and AI—To be Bridged by 'Responsible Governance'

Mr. Marwala first shared the fact that progress on the Sustainable Development Goals (SDGs), adopted by all UN member states in 2015, has stagnated.

'Only 18% of the targets are on track to be achieved by 2030.' (Mr. Marwala, summary)

He then mentioned the potential of AI.

'Research shows that AI has the potential to support the achievement of 79% of the SDG targets. It can provide innovative solutions to the most complex challenges facing humanity.' (Mr. Marwala, summary)

However, Mr. Marwala warned that this potential would not be unlocked unconditionally. He stated that the irresponsible use of AI could undermine the progress of the SDGs themselves and continued his discussion on the main theme of the lecture—AI governance.

2. Four Elements that AI Regulation Should Target

Mr. Marwala explained that the term 'regulating AI' can actually be broken down into four different targets.

- Data: How it is collected, analyzed, stored, and used; cross-border data flows; representativeness of datasets. - Algorithms: Which algorithms to choose, and why; balancing transparency and security. - Computing: Procurement of semiconductors and raw materials; water and power consumption of data centers. - Application: Different accuracy requirements for each industry (e.g., 95% accuracy is insufficient for medical diagnosis).

He then pointed out that a single entity should not regulate all four elements, but that roles should be divided among different layers.

'Which algorithm to choose should be decided by the industry, not the government. Cross-border data flows should be handled by international organizations. Data centers within member states should be regulated by those member states. That is why we need to harmonize these.' (Mr. Marwala, summary)

3. A Warning Against 'Regulatory Arbitrage'—International Cooperation is Essential

Mr. Marwala cited 'regulatory arbitrage' as the most significant structural risk to be aware of in AI governance.

'The situation where AI companies move their bases to places with lax regulations—this is called 'regulatory arbitrage.' Without cooperation at the global, national, and industrial levels, such loopholes will be created.' (Mr. Marwala, summary)

As specific examples, he mentioned the EU's AI Act and GDPR, stating that they serve as good references for regulatory frameworks in other regions.

4. The Strategic Significance of the UN University Being Based in Japan—Specific International Frameworks

Mr. Marwala introduced the specific activities through which the United Nations University (UNU), headquartered in Tokyo, is leading the construction of an international framework for AI governance.

- UNU Global AI Network: A network of over 100 member institutions that coordinates AI policy and research. - Microsoft × LinkedIn × UNU Joint Course 'Responsible AI in the Global Context': Offered free in English and Japanese, with a certificate of completion. - UNU AI Institute (Bologna, Italy): Newly established as a European base. - Global Digital Compact: A UN framework to address the access gap to AI technology and the devices needed to use it. - UN International AI Panel: Creates reports on emerging risks and how to address them. - UN Office of Digital and Emerging Technologies: Oversees not only AI but all digital and emerging technologies.

At the beginning and end of the lecture—from Dr. Hawking to Secretary-General Guterres...