Data and Its Discontents

We certainly still can build these mechanisms to take advantage of AI’s benefits and protect against its challenges, but the window to respond is closing. AI law is still in its infancy. It is essentially at the point of online data privacy law in 1995 — before a series of laws like the EU Data Protection Directive (1995), HIPAA (1996), and Gramm-Leach-Bliley (1999) began setting a somewhat uneven rules landscape for data protection in the digital world. Many would argue that today’s internet ethos is a result of not taking the chance to create a legal and moral structure online more seriously. What lessons can we learn from the early internet age to think about undergirding AI in a way that is built to last? How should that legal framework look to match our values? How can we align and coordinate regulatory models in areas like privacy, competition and fairness to make sure they are reflected in AI governance?

Take issues related to data ethics, one of the direct antecedents to AI ethics. Across the world, data regulation varies distinctly by country, even between liberal democracies. For example, Germany’s approach to personal data protection is based on the Kantian conception of human dignity. The German social contract enshrines this notion of ethics in Article 1 of the German Constitution: “Human dignity is inviolable. To respect and protect it shall be the duty of all state authority.”[i] That first principle serves as the ethical basis for German and European approaches on everything from privacy, like the EU’s General Data Protection Regulation (GDPR), to policy regarding the future of work.

The notion is closely aligned with — but distinct from— the Anglo-American notions of self-actualization. Think about the “pursuit of happiness,” a notion based on the human doing rather than the human being. Perhaps that is one reason why many data protection laws in the US are based on human activities — taking care of your health or managing your finances. These shades of difference have led to some titan battles in the world of tech across the world.

But these differences in liberal democracies remain firmly rooted in the centrality of the individual. That creates a number of conflict points with, for example, the Chinese government, which approaches data governance in a way centered on the notion of social harmony[ii]. These comparative differences have real-world impacts when considering AI research and development, automated decision making, human-in-the-loop authorization and consent, and data usage and safety. Ultimately, it impacts our relationship with technology and power.

What’s more – AI’s concentration in a small number of state actors could have under-examined implications for the Global South. Will they become “AI takers” to new data colonial empires from Silicon Valley and Beijing? What could it mean for countries in Latin America, Africa and Asia if critical media, behavior nudging and disinformation – all of which play important roles in democratic integrity at home – are increasingly operated and regulated abroad? Alternatively, could a lax regulatory environment in developing countries lead to testing laboratories for AI, and what positive and/or negative effects would this have?

[i] The Federal Government. Basic Law.

[ii] Pfeifle, Sam. China’s evolving views on privacy. (International Association of Privacy Professionals, September 28, 2017).

How Should AI Be Governed?

Throughout history, people have adversely reacted to the introduction of new technology, from books and the telephone to cars and the personal computer. While new technologies like AI promise to improve efficiency and competitiveness for countries, citizens, and companies, the negative impacts on society should be minimized through governance. Many participants underscored that regulation and testing are essential to fully understand an AI’s implications before it is rolled out in society. This policy would be harder to implement than to demand from companies and governments. For Germany, quickly introducing AI should require accompanying oversight of the technology by companies and government. Yet, cautiously introducing AI is not always a priority. Central Europe and Mexico, for example, prefer the quick application and adoption of AI in order to be competitive with technology leaders like the US and China.

The challenge of balancing the benefits of AI with fears about its repercussions is not limited to one country alone. All governments would benefit by examining and accurately conveying how AI will change the delivery of services for citizens, and not just as a technology that will disrupt daily life. By dispelling misconceptions and gathering feedback, governments could more successfully adopt AI, and citizens could ensure that the decision-making process is transparent and protects against unethical practices and biases.

Regulation in the EU has been largely successful due to the existing trust in public institutions. Because of this, novel quasi-governmental institutions, as suggested by Germany and Central Europe, could help vet new AI technologies and create policies with companies to ensure the technologies align with core European values. These policies could then be transposed into the national legislation of EU Mem­ber States, which would support consistency and data protection.

An EU-wide labeling system, similar to ecological standards for household appliances or organic foods, could inform consumers about the technology’s creation, contents, and capabilities — including how consumers’ data would be used. Actions such as these would help the EU become a leader to mitigate ambivalence surrounding AI, given that it has some of the world’s highest consumer protection regulations. This idea, however, would not necessarily be applicable in countries without sufficient consumer protections or trust in governmental institutions.

Participants proposed independent advisory boards as a mechanism to include voices from government, technology, and non-governmental institutions. For the UK, these boards could play a key role in shaping decisions on best practices and systems to regulate new technologies. Similarly, in France, incorporating feedback from groups adversely affected by automation, disenfranchisement, and inequality could inform the development of AI to ensure that technological innovation is tied closely to social impact and mitigates the negative disruptions to society.

Establishing rules to design and develop AI are important, but participants in Central Europe recommended the region avoid over-regulation to prevent stifling innovation and competitiveness. The only country for which participants suggested governing AI through an international lens was India, proposing that the Universal Declaration of Human Rights be an anchor to address governance of ethical AI, given the socioeconomic challenges unique to the Indian context. It could be that each country develops its own AI systems with little regulation, or where countries export AI technologies to others in a fair and ethical way. From these discussions, it was apparent that an international solution to govern and regulate AI would not be politically feasible, or necessarily desired, for all the NextGen countries.