Framing the issues: the evolving state of AI governance

Author: Yuko Harayama, Professor Emeritus, Tohoku University 

Furthermore, with the rise of generative AI, our relationship with AI has changed drastically, since we are now directly using AI as a personal assistant or “partner” via quasi-conversation. You can ask any question on any topic by using a prompt, and you receive almost instantaneously a response, accurate or not, within the parameters fixed by the developers of the system. You can initiate a conversation on any issue, then you have “someone” reacting to you and expressing an “opinion” related to your thought. Also, AI, in particular generative AI, has become a topic of conversation not reserved for software engineers or computer scientists, but for ordinary people. This is where we are today. 

What makes AI so different? 

Throughout history, humans have created technologies to ease daily tasks, to improve the quality of life, and to challenge their own capacities. And technologies, in their turn, have shaped the structure and the functioning of human society, for better and sometimes for worse. AI will be no exception, but the difference from other technologies lies in the speed, the scope and the depth at which AI is impacting society and the values that underpin it. This may imply that societies are left with limited capacity to react once harm is done. A further characteristic of AI is its inherently dual nature. AI, once developed, can be deployed in different contexts with almost no additional effort, potentially diverging from developers’ good initial intentions and causing harm to humans or threatening public security. 

Responsible and ethical AI 

Therefore, most governments advocate for responsible and ethical AI, and recognize the need to develop an appropriate AI governance framework and to accompany the advancement of AI with so-called guardrails, before it is too late. International organizations have already taken concrete actions in this direction, starting with the OECD’s AI Principles (2019) based on its human-centered approach, and UNESCO’s adoption of its Recommendation on the Ethics of AI (2021). In a similar vein, the European Union has launched the Ethics Guidelines for Trustworthy AI (2019) prepared by its High-Level Expert Group on AI. 

Following these first moves toward AI governance, international, regional and national efforts are underway to accompany the implementation of these principles (e.g., the OECD AI Policy Observatory, UNESCO’s Readiness Assessment Methodology and Ethical Impact Assessment), to address some specific challenges, such as AI safety, or to translate these principles into actionable and implementable tools. For the latter, the Global Partnership on AI (GPAI), aiming to “bridge the gap between theory and practice on AI” was launched in 2020, and the United Nations convened a multi-stakeholder “High-level Advisory Body on AI” in late 2023, which recently published its “Interim Report: Governing AI for Humanity”. Also, within the framework of the G7 Summit under the Japanese Presidency, the so-called Hiroshima AI Process has been initiated, resulting in an “International Code of Conduct for Organizations Developing Advanced AI Systems”. 

On the other hand, the EU expressed its preference for a more binding tool. The European Commission proposed an AI Act, founded on a risk-based approach and proportionality in 2021, which has been discussed and voted for at the European Parliament in June 2023. 

At the national level, most AI governance initiatives are oriented toward the implementation of guidelines and the creation of an oversight and advisory body. To name a few, US government has established the National AI Advisory Committee, Canada is equipped with different tools, such as its Algorithmic Impact Assessment and Guideline on Service and Digital, and Japan is preparing its guidelines for companies related to AI. 

Where do we go from here? 

This quick overview of the global trends on AI governance shows that an ever-increasing number of initiatives is underway, while their approaches, the sphere of influence and the level of enforcement vary depending on the political and socio-cultural context. This multiplicity and multi-layer structure of AI governance could be a barrier for the advancement of Responsible and Ethical AI, and given that the development and use of AI transcend national borders, we need to encourage more coherent and coordinated actions among nations and international organizations. In this perspective, the approach chosen by the Committee on AI (CAI) of the Council of Europe in its effort to elaborate an AI Convention based on human rights, democracy, and the rule of law, is insightful. 

With the ambition to make the Convention “a global instrument”, signifying that it could be ratified by many non-member states, the CAI allows observer states, which include the US, Canada, Japan, and Israel among others, to participate in the drafting of the Convention. Also, in its effort to reach an agreement on the fundamental principles, recognizing the diversity of legal systems, the CAI is demonstrating openness to find compromises and innovative solutions. 

2024 will be rich in terms of technological advancement in AI, new products and services created using AI, and creative applications of AI, while the downsides of AI persist. More challenges are to come for ensuring responsible and ethical AI governance. 

_{The ideas and opinions expressed in this article are those of the author and do not necessarily represent the views of UNESCO. The designations employed and the presentation of material throughout the publication do not imply the expression of any opinion whatsoever on the part of UNESCO concerning the legal status of any country, city or area or of its authorities, or concerning its frontiers or boundaries.}