Introductory Remarks by Michael Cooper

I am the Executive Director of the American Society of International Law. It is my pleasure to welcome you to the 119th annual meeting and also the Tenth Vagts Roundtable. So I hope that you are all members of the Society, and if you are not, I trust that you will be, by the end of the day, loyal members.

One of the things I like to point out to members, especially the younger ones, is that we do have a special program where after your fiftieth year of membership, we waive the membership dues from that point forward. Believe me, if you are young, fifty years will pass. It is a good time to jump in now, and you can look forward to those free membership dues down the line.

I would like to start out by acknowledging Detlev Vagts, who is also known as “Det,” for his immense contributions to the field and his longtime dedication to this Society. By the end of his career, among his many contributions, Det was known as one of the great framers of international law in the context of transnational law, which is vital for us to solve transnational problems. We know that there are many of those problems today. Those who knew Det speak of him as a supreme gentleman, as an exceptional scholar, as a generous, insightful, and inspiring teacher, and for us here at the American Society of International Law, we could not be more pleased than to honor his life and his memory, including through this roundtable, once again celebrating its tenth year.

I want to thank Karen and Lydia Vagts and the whole Vagts family for their generous gift that has enabled the Society to create this annual event year after year, and I also want to thank our convener, Professor Marc Rotenberg, for pulling this panel together this year. He is the founder and the president of the Center for AI and Digital Policy. He is also on the faculty of Georgetown Law, my alma mater. I would like to thank you, Marc, for putting this fantastic panel together.

It is my great pleasure to invite you, Marc, to open up the Tenth Detlev F. Vagts Roundtable.

Introductory Remarks by Marc Rotenberg

Great. Thank you, Michael, and thank you to the American Society for International Law for hosting this event. And thank you to our panelists for being with us today.

For those of you who are not familiar with the Vagts Roundtable, we will have a presentation by a leading scholar and then commentary by respondents, and this is also to honor Detlev’s tradition of mentorship and recognition of young scholars in the field of international law. I could not imagine a better person in a critical field of international law to speak with us today than Nathalie Smuha. Nathalie has emerged as one of the world’s leading experts on AI, artificial intelligence, and the law. She is on the faculty at the New York University School of Law, KU Leuven. She has been an expert advisor to many international organizations, including the Council of Europe Committee on AI, the OECD AI and Data Protection Group of Experts. She worked in the early days on the development of the EU AI Act with a high-level expert group that actually provided the insight and the foundation for what is today the world’s most comprehensive regulation of AI.

I am going to introduce our panelists after she has presented her paper, which is titled “The International Law of AI: A New Legal Field?” I will note that she ends her title with a question mark. I did a bit of research about Detlev’s legal contribution, and I found that he was well-known for an article that he published, “The International Legal Profession: A Need for More Governance?” also ending with a question mark. But the former president of ASIL, Lori Fisler Damrosch, remarked, “Although the title carries a question mark, there was no question in Detlev’s mind that the need for transnational professional ethics was profound and that this Society should take up the challenges of beginning to fill it.” I suspect that Nathalie, in asking the question, as for the existence of the field of AI law, has an answer for us as well.

Over to you, Nathalie.

Remarks by Nathalie Smuha

I have to start by thanking all of you, and especially Marc, for the invitation of presenting a paper here today. I have been wanting to write this paper for a very long time. I just did not have the opportunity to do so, and it helps if someone invites you with a strict deadline to deliver a paper. I also think this is the right time to write a paper on AI and international law because we have been having so many developments in the field lately.

There are many questions one could ask about AI and international law, and a growing body of scholars around the globe are asking these questions. To give some examples, how is international humanitarian law applicable to AI in armed conflicts, for instance? How do international organizations deal with the rise of AI policies around the globe? What kind of legal rules are being adopted on AI around the globe? And should there be regulation of AI at the international level?

The question I am asking in this paper, however, is slightly broader and a little bit more meta. I am asking, what if the international law of AI already exists? How would we know? And what would be the implications of that?

To conduct this inquiry, I explore five questions in the paper. The first one is, what would the international law of AI actually look like? How would we recognize it? The second is, what kind of criteria can we use to examine and identify its existence as a potential legal field? What makes a legal field a legal field? The third question is, to which extent are these criteria present in the context of AI? To which extent could we conclude that today there is already a self-standing legal field, the international law of AI? The fourth question is, if it exists, what are the implications thereof? And then finally, what would the future of the international law of AI look like, regardless of whether we acknowledge it is a new legal field? In the next ten minutes, let me take you through each of these five questions.

First of all, what would the international law of AI look like? How would we recognize it? To answer that question, I am breaking down each of the components of the international law of AI. Starting with international, international law, we all know, is clearly different from domestic law. There are some crucial differences, like the lack of a central enforcement authority but also the lack of a central enactment authority. And there are specific sources to international law. These are listed, for instance, in Article 38, Paragraph 1 of the Statute of the International Court of Justice. We would imagine that if there were to be an international law of AI, the sources that we would find it in are international conventions, treaties, customary law, general principles of law, and subsidiarily, judicial decisions and teachings of—and I quote the article— “the most highly qualified publicists of the various nations,” whoever that may be.

One aspect that is not listed in Article 38—but it is very important and increasingly important in the field of international law—is soft law. I also examined to which extent soft law in international law is already covering this field.

What about law? What kind of legal rules could we imagine in an international law of AI? Law is a very versatile tool, and it can have many different purposes. When we think of the law of AI or regulation of AI, typically what comes to mind is a restriction of certain AI practices, maybe dangerous or risky practices, but law can also be a tool to facilitate the development and the uptake of AI. Think of a law that is adopted by a country to facilitate immigration of AI experts and make it easier for them to come to the country. Or think of a law that mandates subsidies for companies that develop AI. These kinds of things, which facilitate the technology, you also adopt through laws. Legislation can also serve as a tool for that.

And then countries need to decide, will their laws be binding or non-binding? Will they be technology neutral or technology specific and focus only on AI? Will they impose obligations on developers of AI systems, ex ante, exposed? Will it be a risk-based approach, a rights-based approach? What will fall in the scope of the law? What will fall outside it? Think of defense, national security, often excluded.

The third element of the international law of AI is AI. Originally, this was seen as a scientific discipline that researchers were focusing on. Only more recently, with the uptake of the technology in all domains of society, did it become a noun. I am using an AI to translate something, for instance, or to chat. You can also find very hyperbolic statements, like a report from the European Parliament a few years ago calling it the “fifth element after air, earth, water, and fire.”

There is a whole spectrum there of how people view AI, and its precise contours are a gray area. When is a technology intelligent enough to be called artificial intelligence? And when is it just a computer system and not worthy of the name AI?

The fact that it is a very vague term has led some people to conclude that you cannot regulate it. But that would be a mistake, right? Because first of all, lawyers are very good in defining things. We are all about creating definitions in legislation and then regulating it, and so there have been many initiatives of defining AI system and what it means so that we can subject it to policies.

One of the most influential definitions comes from the OECD, so an international organization bringing member states together to reflect on how we could define this. In Europe, the High-Level Expert Group on AI was one of the first that put forward the definition, and these definitions now tend to converge, which is an interesting point.

But alternatively, instead of defining AI system, you could also think of breaking it down into its components, because an AI system essentially consists of algorithms and data, and there is already a lot of regulation of data. It is also computational hardware and infrastructure. Think of CPUs and GPUs, chips. These are all things that come to mind much easily controllable through legislation. I think the idea that we cannot regulate AI because it is this thing that we cannot grasp is a mistake, and we are seeing that there is ample AI legislation out there already.

The second question, what makes a legal field a legal field? If the international law of AI were to exist, what would be the criteria based on which we could identify it? In my research, I discovered there is no neat list of criteria that a legal field needs to meet in order to say, okay, now this is a self-standing legal field.

Instead, what I did is I explored what are criteria factors that legal fields have in common. I looked a little bit at that, and then I also looked at a very new legal field in international law, relatively speaking, and that is international environmental law. I tried to assess what are factors that we see in that legal field that kind of contributed to it today being seen as a self-standing legal field.

In the end, I decided to look at seven factors that I think are not exhaustive. It is possible that there are other relevant factors, and it might also be that these factors in and of themselves do not lead to the conclusion that there is a new legal field. But I thought these were good starting points.

A first factor would be the emergence of a new or a newly understood societal phenomenon that seems to require regulatory intervention at the international level, which is something that is, for instance, very strongly seen in international environmental law. A second factor I explored is the pertinence of existing international law and the need for further guidance as to how to apply that international law. A third factor is the adoption of subject-specific instruments, distinct legal texts that deal with the topic. A fourth factor, the implementation of national legislation that has a transnational impact and also global regulatory convergence on national legislation on the topic could also be elements pointing towards an international legal field. Fifth, for a legal field to be a legal field, you typically have the emergence and development of theories and doctrines that underlie these subject-specific legal texts. Sixth, you typically also have specialized actors, legal scholars that focus on a specific topic or practitioners that focus on a specific topic, and sometimes you even have dedicated authorities and procedures to govern a legal field. Finally, I also looked at the development of global standards by international standardization organizations, which is something you also see in the field of international environmental law.

I am coming to my third question that I explored, and that is, these seven criteria that I just listed, to which extent can we find evidence of their presence in the context of AI and especially the international law of AI?

I think everyone would probably agree as to the first factor that yes, there is the emergence of a new societal phenomenon that seems to require regulatory intervention at the international level. The effects of AI systems do not stop at the national border. If a Country A develops and uses a technology of an AI system, the negative effects could also spill over to Country B, C, and D. So you would need some sort of international cooperation to deal with that. There are negative externalities, and if states do not cooperate, they might not be tackled. This is not only with the use of AI, but also, if you are going to look at the development and the value chain of the development of an AI system, it is very international too.

As to factor two, we see that there is already international law that applies to AI today. It is very pertinent, but guidance might be needed as to how to apply it. Think of human rights law. International human rights law is extremely relevant for the field of AI. Rights like privacy and equality, they apply to AI. They are important for AI because they can be at stake when AI systems are used, and we see that international organizations are developing guidance as to how these bodies of law should be applied to AI more specifically. I think this criterion might be fulfilled as well.

What about the adoption of subject-specific instruments of international law? Do we already see legal texts in the international legal arena that focus specifically on AI, rather than technology in general? We also see evidence of that. The first international treaty on AI is a fact. It was adopted by the Council of Europe last year. The treaty has more than ten signatories as of yet. The United States is one of them. It was negotiated by not just the Council of Europe member states, but also states like the United States, Canada, Israel, and others. It does seem to be an international convention. It is open everywhere across the globe.

It is not in force yet. We do not have ratifications yet, because it was only open for signatories in September last year, but it is a binding treaty of international law specifically focusing on AI, human rights, democracy, and the rule of law.

The European Union, of which the status of an international law organization is a bit contested, nevertheless came up with an Artificial Intelligence Act, which binds its twenty-seven member states. When you look at soft law initiatives, the African Union, for instance, adopted a continental AI strategy, and numerous international organizations adopted recommendations and guidelines on AI. There is a growing body of hard and soft international law of AI already.

As to the fourth factor, the implementation of national legislation with a transnational impact, many countries have adopted soft law instruments to guide the development of AI in a responsible manner and even more when it comes to the uptake and promotion of the technology, but also increasingly hard laws. China was one of the first countries to adopt AI-specific legislation. South Korea adopted legislation. Brazil is in the process of doing so. In the United States, some states have been legislating AI, such as Colorado.

Interestingly—and this brings me to the fifth element—there seems to be a growing number of theories and doctrines underlying these separate initiatives, because in almost all these initiatives, you see that the main aim is to boost, maximize the benefits of the technology while minimizing the risks. This is literally the formulation you can see in the documents of the EU and the African Union, for instance, very similar language. Very often you see similar terms coming back there: transparency, accountability, human oversight, and control. There seem to be some general principles that we can start distinguishing in this body of international laws of AI.

You also start seeing specialized actors. Law firms start having an AI practice and not just a privacy practice. In legal education, you can find courses now on artificial intelligence regulation. I am teaching a course like that, and almost every university now offers these kinds of courses.

Across the globe, you also start seeing specific authorities that focus on enforcement related to AI. In the European Union, we have the European AI Office. In the UK, you have the AI Security Institute. The OECD has an AI policy observatory. The Council of Europe Convention will have a conference of the parties. Lastly, we also see ample development of international standards of AI; ISO standards, for instance, or IEEE. NIST is developing standards in Europe since CENELEC. To conclude, it is not unreasonable to say that such a legal field emerged.

I am going to close by focusing on why this discussion matters. Is it just a matter of legal taxonomy? I do not think so. There are potential downsides of starting to recognize legal fields just like that. There might be proliferation of new legal fields. In the 1990s, when the internet appeared, Frank Easterbrook wrote a famous article on the law of the horse, saying that, well, we should not have law of the horse topics taught to our students because we are going to lose generality. There are many different bodies of law that apply to horses and horse sales, et cetera. He might say something similar about AI.

I think AI is different, and I am not alone in thinking that. But it is true that we should not lose sight of the many existing bodies of law that also apply to AI outside of AI-specific legislation.

I do want to draw attention to upsides, benefits of recognizing this legal field. Recognizing that there is a self-standing legal field helps us to understand what it is, what it does, what its underlying purpose is. It would also allow us to delineate it and hence to critique it and to uncover potential legal gaps. Having a distinct legal field also helps us to establish a common vocabulary across the globe to pinpoint where things can go wrong, where we need to understand each other and cooperate with each other. Acknowledging it can also help encourage international and domestic actors to see, okay, there already is an international law of AI, and we need to implement these rules that apply to it. We cannot just ignore it, which can lift the level of protection. It can also offer counter-narrative against the nationalistic turn we are seeing across the globe.

Which brings me to my closing statement, and that is the fact that international law, not only of AI, but in general seems to be in peril. The level of willingness to cooperate internationally seems to take a hit. We have seen states breaking with international courts or even threatening to take legal action against their staff, refusing to comply with court judgments, withdrawing from international organizations, engaging in state actions that undermine international cooperation, launching trade wars, to give but a few examples.

This disengagement with international rule of law might also have repercussions for an international law of AI. This is something we need to be mindful of, but I would conclude that this means now more than ever, it is important to develop an international law of AI because there are important risks links to this technology which require international cooperation.

Thank you very much.

Marc Rotenberg

Thank you, Nathalie. That was a wonderful presentation, gave us all a lot to think about. We will have a response from expert commentators and then I think an opportunity for those in the audience to ask a few questions as well.

I would like to introduce our first respondent. Tammy Horn is Senior Global Regulatory Policy Counsel at IBM for AI and Data Privacy. She has had senior roles at IBM for over eighteen years. Her portfolio over this time has covered a wide range of issues from cybersecurity, encryption, and government access to data, to the GDPR, export controls and economic sanctions, federal ethics and lobbying laws. So that is quite a range of activities. She is a former associate of WilmerHale here in Washington, D.C., and we are very pleased to welcome her to the ASIL conference. Tammy.

Remarks by Tammy Horn

Thank you for including me in this distinguished panel. It is great to be here, and let me just start with the requisite disclaimer that my words are my own and not attributable to IBM.

I am happy to share my perspective as a practitioner in-house counsel at a hundred-plus-year-old technology company that is committed to building and using technology for good, including trustworthy AI. I have to say that reading your paper, I felt like you have been with me for the past six to seven years along this journey as we navigate within IBM, all of the explosion of activity in this space of regulating AI.

Even before there were AI laws or AI safety institutes or codes of conduct, IBM was thinking about living these issues that you raised in your paper. We recognize the incredible potential that AI represents for the world to help solve the world’s problems, but we also recognize the incredible responsibility to develop and deploy AI solutions ethically and to protect against the risks that AI presents.

To this end, we were one of the first AI companies to establish an AI ethics board that comprised a multidisciplinary team, which I think recognizes the fact that there are lots of different areas of law that interrelate with AI, and that ethics board is part of a larger kind of integrated governance program and framework within IBM that drives ethics by design and how we manage our internal governance.

Some of the challenges that we have faced over the past couple of years is navigating this emerging regulatory landscape, saying, okay, is this a new requirement? Is this something that we already address through our GDPR and privacy governance framework, or is there something new that we need to do? I agree with your premise that there is a current field of AI law. I have AI in my title. Many people do within the law department at IBM, but then someone else may work on something related to AI, but they do not have AI in their title. Where do you draw those lines and how do you define it?

At IBM, we have been actively participating in and shaping some of the regulatory developments, including at the OECD. We were actively involved with developing the definition of AI, the United Nations, various regional multilateral codes of conduct, the international standards work, and navigating this explosion of AI laws and frameworks. What we have seen is that many of these things align, but a lot of times, they also are different, but I think they all show evidence that there is a field of international AI law.

There also are emerging laws and regulations in the areas of privacy, cybersecurity, national security, cloud, competition, the long list of things that you mentioned that I cover in IBM, and they all impact and touch on AI in some way, shape, or form. Your paper resonated with me. It raised questions that not only are intellectually interesting but also relevant to how do you navigate this complex web of emerging laws and regulations and standards and best practices and codes of conduct that either are directly or indirectly regulate AI.

There is an analogy to the issues you raised in your paper about whether the field of international AI law exists or should exist and how it should be defined to the experience of navigating this emerging regulatory landscape in the context of a global technology company.

Addressing questions like, is AI its own separate field of law? Where does it begin? Where does it end? What do you do with the overlaps with laws like anti-discrimination, employment, human rights, privacy, cybersecurity, national security, competition? How do you account for those? This is something that even within a multinational company, we navigate every day to try and address how do we define it, how do we make sure that we are doing the right thing, how do we incorporate this into our internal governance program.

It raises questions that can we rely on these existing areas of law? For example, for anti-discrimination, it already applies to the use of AI. Do we need a separate AI law to address discrimination? We have national security laws that restrict the use of AI for certain purposes or restrict the export of certain technologies to countries of concern. Do we need a separate AI law to address those issues, or can we rely on what already exists?

This brings to light the reality that we need a thoughtful, multidisciplinary, and well-organized approach. From a practitioner perspective within a multinational company, navigating duplicative or inconsistent requirements can be very challenging and burdensome, and so creating duplicative and inconsistent requirements under a new AI legal regime could be confusing and burdensome and undermine widespread adoption. If at the end of the day, what we want is widespread adoption in order to address the risks and protect against the harms, we have a responsibility to be thoughtful in the approach of how we define AI law and how we integrate all of these other areas of law that touch upon AI.

There is an argument for relying on existing norms, standards, and laws rather than standing up a whole new field of law like the law of the horse, as you referenced in your paper. This approach also would address the issue of relevancy, that you do not want to create a law that becomes obsolete because the technology is evolving so quickly.

But, on the other hand, there are many issues that are unique to AI, many risks that we need to think about and that necessitates a more comprehensive and interdisciplinary approach across the relevant legal domains. I do not think this has to be a binary choice. I think that the field of international AI law could be viewed as a comprehensive, self-contained framework that both integrates existing laws, regulations, and standards, and also addresses any gaps to address the risks and opportunities unique to AI.

I look forward to seeing more of your scholarship and work in this space, and I really enjoyed reading your paper. It resonated with a lot of the issues that we have been dealing with within a multinational company. Thank you.

Marc Rotenberg

Thank you.

Our next speaker, Catherine Régis, is a full professor at the Faculty of Law at the University of Montreal. She is a co-director of the Canadian AI Safety Institute Research Program and director of Social Innovation and International Policy at IVADO. In addition to holding a Canadian CIFAR Chair in AI and Human Rights and a Canadian Research Chair in Health Law and Policy, she is a Senior Research Associate at the Intellectual Forum of the University of Cambridge. Catherine has provided expert advice for many international organizations, including the OECD, the World Health Organization, and the United Nations. Thank you for being with us.

Remarks by Catherine Régis

Thank you very much for the invitation. I am super happy to be here, and I want to say congratulations for your paper and your scholarship on this, Nathalie. I think we need it now. It is a fantastic contribution, and I invite you all to read the paper.

Something interesting that your paper provokes in me is kind of a therapy and it made me think about my own journey in the legal domain. I will explain why. In the paper, you explain the different steps in creating a field of law and the criteria. I was thinking about my own journey in the legal domain and where I am right now and what kind of reflection are inhabiting me at this moment, a key moment in AI governance.

My first phase is when I was a legal student—I will call this “idealistic phase.” That was the time when I thought law was a very shiny object, and law was the solution to all the problems. That was the first phase. I was excited and optimistic.

Then I became a lawyer. That is my second phase. I will call it the “pragmatic phase.” Now you realize that all the fantastic information or knowledge that you receive at university is clashing with the reality of the practice, and then you realize that there are many challenges in applying the law. We can all think about different challenges, but for me, one that stood out, it was access to justice for a lot of people.

I was thinking now as a legal scholar, but also someone who works on different projects on AI governance, where am I now? I would call this the “integrative phase.” What I mean by that is I realize now, at least based on my own experience, that law is only one element, but a very important element of a broader normative strategy that include various tools. Law is evolving with standards, and you mentioned that component, Nathalie, but also financial incentives. We have research incentives and institutions, and it is a broad portrait that influences how we manage this technology.

Because of this different phase and where I am right now, I think the questions or the comments that I would like to make are organized around the normative playground that we have and how law is one component, but only one component of that broader strategy of our normative playground.

The first comment is, in the context of an evolving field, there are a lot of actors, including policymakers and professionals like physicians and regulatory colleges that ask the question, how should we best regulate AI? What is the key way to do this? It is a very difficult question. The AI Act is an illustration of organizations who took a stand on this, but there are other people or other actors around the world who are still asking the question.

The question I would have for you, Nathalie, is people who ask this question, what would you say about the contribution of the law, knowing that there are other normative strategies that can be influential as well? We talk about standards, but there are others, and knowing that it is a complex puzzle where law plays a role, but it has to play with these other norms. Do you think that we should train a new generation of lawyers who will have a key role in AI governance to be able to understand the role of law in combination with these other normative components? Think about how we should develop the law in a context where we can play with these other tools. It is not only about being super targeted about the contribution of the law, but how it can play with these other normative strategies.

It is very difficult to do in practice. When people ask you what we should do and we all recognize that there are different tools in AI, it is such a complex strategy, how do we mobilize the tools, and how can we make a coherent puzzle? Should we develop that kind of thinking in not only legal scholars, but also legal students and lawyers? That is my first question.

The other one, I like your discussion on the contribution of soft law compared to hard law, and we often see this as a binary structure. It is one or the other, or they are in a position.

I have been doing a lot of work on the WHO and how the norms travel from the international level and domestic law, and there are many parallel lessons that we can derive from that kind of work for AI.

I have been studying this for five years, the circulation of the norms. I studied more than 5,000 references in domestic law to international norms, either hard law or soft law, and what I realized is that they both find their way in domestic law.

In the context of the WHO, the World Health Organization, actually soft law is often more cited by courts in regulation and in legislation. Then it made me question, is it still relevant to see this normative contribution from the international level with this binary distinction? Are there other criteria that give us more understanding of why a norm would be taken into domestic law? In the research I conducted—and it is also coherent with some other researchers in other international law field—we realized that it is sometimes not about this distinction between soft law and hard law, but more about some characteristic of the norm in itself.

One of the key characteristic—I think it is interesting to see whether it will have influence on the domestic level—is the technical or the scientific component of the norm. When there is a strong technical and scientific component in the norm, for instance, how much level of noise we accept in a neighborhood, so people will not be stressed or disturbed by that kind of noise. It is very precise. It is very technical. It finds its way easily in domestic law. Compared to broader reference to norms or values such as your norms at the domestic level should make sure that everybody is treated equally, that it is more difficult to find its way at the domestic level.

I want to say about the soft law versus the hard law that an interesting body of research also found analyzing hundreds of treaties, that one of the key elements that will make a difference in the domestic law is accountability mechanisms. It is not necessarily you have soft law or hard law but whether any of these mechanisms or norms have accountability mechanisms that make sure that people will comply with them.

Marc Rotenberg

That is great. Thank you so much.

I want to mention, by the way, the work of our expert commentators is available online. They have extensive publications, and I would encourage you to go online and see the contributions that they have made in this field.

Nathalie mentioned earlier that one of the key domains of international AI law would be the field of human rights, and I could not imagine a better person to speak with us today at the annual ASIL conference than John Shattuck. John is professor of Practice and Diplomacy at the Fletcher School of Law and Senior Fellow at the Carr Center for Human Rights at the Harvard Kennedy School. He has had a distinguished career in law and diplomacy and education. He served as Assistant Secretary of State for Democracy, Human Rights, and Labor for President Clinton. He was Vice President of Harvard University and President and Rector of the Central European University. He served as U.S. Ambassador to the Czech Republic. He is the author of three books, including “Freedom on Fire: Human Rights Wars and America’s Response.” I should add, by the way, he was one of my mentors and got me interested in the field of technology and law. I am absolutely thrilled to have John with us today.

Remarks by John Shattuck

Thank you, Marc. Of course, you are one of my great achievements. Thank you so much, and congratulations to CAIDP, which Marc founded and heads and has been instrumental in gathering information and developing policy recommendations on AI.

I do want to congratulate Nathalie very much on this remarkable paper because it is such a broad topic, and you wrestled it to the ground with great elegance. You did that very well in your remarks. Thank you.

I am going to focus on only the last very short remarks that you made about the future, and to do so, I hope I do not turn myself into the skunk at the picnic. But I will try to be gentle about this. This is not in any way to reflect negatively on the paper. The paper was superb, but it is to look at the rapidly changing geopolitical context in which this new field is being developed and the incredible pace with which it is being developed, as well as the pace of geopolitical developments that are affecting it. I want to look at four big trends. You will recognize all of them, and they are impediments to the future development of the international law of AI, which I will suggest can be removed or at least addressed in certain ways, but it is a big task.

First is the rise of authoritarianism or the rise of anti-democratic activities inside democracies, to say nothing of authoritarianism over in other parts of the world, which are non-democratic. A huge pushback—and you referenced this, Nathalie—against the rule of law, the rule of international law, the rule of law domestically, and the way in which law is increasingly seen as an impediment to power, and that itself is a very disturbing development for a field such as international law.

This is, to some extent, fueled by the populist revolt against various things but particularly against the economic and cultural globalization that we have seen over the last twenty to thirty years. I will give you one example of the pushback against the rule of law by power—Elon Musk and his DOGE team have obtained access by all reports to the Social Security information of millions of Americans, notwithstanding the fact that that information is protected by no less than five pieces of legislation in the United States, including the Privacy Act of 1974, which I worked on back in ancient history. That is point one.

Point two is the rise of nationalism coupled with authoritarianism and the emergence of what we see as great power competition, the competitive race for national dominance, and it is going on, especially right now, between the United States and China and being played out in a variety of different ways. But AI is, in some ways, at the center of that, whereby the United States and China are seeking to dominate the field of AI development, and in the case of the United States, looking at innovation as the most important aspect of any element of law related to AI. A concern that other countries, which are much more oriented toward law, particularly in Europe, where the regulations and new AI law that Nathalie referenced have been developed, those parts of the world, democratic parts of the world, are being left behind in this great power competition between the powers that are seeking to dominate the innovation in this field.

The third element that I would see as a negative geopolitical climate is the rise of an oligarchy and a strong oligarchy, which is increasing in its political power, especially as it works together with authoritarian or anti-democratic leaders, and also works to gain control over the regulatory systems in some ways. Those are very general remarks, and obviously they do not need to be looked at in great detail, but certainly, this is a trend that needs to be identified and worked through. I know Nathalie is going to do so in the next phase of her paper.

The final trend that I would cite is the rise and imminence of artificial superintelligence, or AGI, which we all know is the ultimate prize in the competition, particularly the great power competition that is going on, and who can control the superintelligence that is coming, and, insofar as it is an authoritarian state, it gets used particularly in the interests of pursuing that authoritarian power. Even if it is governed in theory, as I understand it, the science is beginning to indicate that AGI might well be able to slip out of the regulatory framework that gets developed by virtue of its own capacity to overthink and think beyond what is been developed.

These are the impediments. What are the quick strategies that need to be followed, to leave this on a much more upbeat note, if you will? One strategy, the first and perhaps most important, is exactly what Nathalie has pointed toward, which is we have this incredible body of international law, particularly in the human rights field, which is highly applicable to AI development and is much more developed than it was when Marc and I were working together at the ACLU many years ago to develop pre-digital privacy regulations for data banks, et cetera. The development of AI law is something that needs to be continued and pursued. The next phase is enforcement wherever it is possible to seek enforcement. Enlightened companies like certainly IBM and other companies that have the capacity to enforce in this field obviously are going to continue to do so.

The second level of strategy is more political, and that is to build a constituency for AI governance. By that, I mean a popular constituency. This cannot be simply a theoretical matter that is left to the lawyers. But there needs to be a growing public demand for the kind of governance that we are talking about, and indeed, the polling data shows quite well that Americans and Europeans, in particular, are seeking much greater control in a broad sense—they do not have specific ideas—over AI in a climate that does indicate that it would be possible to develop that kind of constituency. We should remember that there have been cross-partisan alliances in the past that led to the development of privacy law, particularly the privacy law that you cited, Nathalie, that became the human rights law.

Back in the 1970s, 1980s, and 1990s in the United States, there was a coalition of Democrats and Republicans in the United States who actually worked on some of the early stages of the legal framework that you are talking about here. We need to redevelop that. This cannot be a partisan matter. It should not be a partisan matter. All the normative aspects of AI regulation are shared by people who are conservative or liberal. There is no particular breakdown.

The final point that I would make is that we need to create a sense of urgency about the regulatory or the governance process that Nathalie has very eloquently set before us. The urgency can come from exactly what I am talking about, particularly in the prospect of the development of a superintelligence coming beyond at the next level of AI development, which will need to be controlled and governed on the front end before it gets away from the regulatory process.

I would end by remembering—thanks to you for citing it—the international law of human rights that developed so rapidly and so powerfully in the years after the Second World War, developed out of the sense of urgency that came from the Holocaust, the massive number of people killed in that war, and the lack of any international framework to be able to address that. In a sense, we need to reconstitute the mood that existed in the world in the aftermath of the Second World War, without, one hopes, encountering the same factors, i.e., a great war that would lead to that.

Thank you, Nathalie, and thank you for this opportunity to not hopefully to be a skunk at the picnic but at least to put before you the political realities of dealing with this issue. Thanks.

Marc Rotenberg

Thank you, John.

We have just a couple minutes left, and I know that we need to end on time at the top of the hour. The commentary has been so excellent and so insightful, and I know, Nathalie, you will take these suggestions on board as you are preparing your paper for final publication.

I do want to just briefly say a word about the Center for AI and Digital Policy. Actually, both Nathalie and John serve on our academic advisory board. We are a global network of AI policy experts and human rights advocates in now more than one hundred countries trying to take up the challenges and the strategies that John just outlined.

We have two significant publications that may be of interest to you if you are attending this session. We publish annually the AI and Democratic Values Index, which is an attempt to assess the policies and practices of countries around the world for AI. We rate them and we rank them against twelve metrics, thinking about democratic values broadly understood, and we have also recently published a new book, the AI Policy Sourcebook. I actually did this about six years ago. It was the first compendium in the AI policy world of governance frameworks and policy resources. We have now updated it because of the EU AI Act and the Council of Europe AI Treaty. Both of these publications help complement the work that Nathalie and others are doing to help us understand the emerging field of AI and the law.

With great appreciation for all of your participation as well as those of you who are joining us in the room and online, I think this moment where we are in terms of the politics of democratic governance, the increasing significance of artificial intelligence, and the need to find a path forward that ensures that we are able to manage this technology consistent with our political norms makes this session very timely and very important. Thank you again to Michael and ASIL and the Vagts family for making this possible. Thank you.