“There is an old saying that victory has a hundred fathers, but defeat is an orphan.” – John F. Kennedy

Without governance¹, its full potential might not be realised and opportunistic AI businesses, data brokers, organisations and governments might exploit the natural world for their selfish or individual gain. However, the presence of regulations doesn’t eliminate exploitative tendencies but only limits their extent and creates a benchmark for accountable and responsible AI. Also, rule-making and governing, in general, aren’t givens but products of the active participation of all people, voters and residents in various democratic processes. AI technologies and regulations aren’t introduced in a vacuum but in existing social, technological and political ecosystems where they must seamlessly integrate without aggressively disrupting the status quo. The seamless integration of AI is impossible if there are no rules governing its development, deployment and implementation. Therefore, inclusive AI regulations must be implemented swiftly to ensure a smooth and gradual transition into a sustainable AI-driven future.

At this juncture, where a handful of countries are aligning with AI regulations while the majority aren’t, there are two main issues this essay aims to raise that threaten AI’s sustainability. One is the lack of political will and failure of political representatives to make legislation. The second is the general population’s lack of substantive representation and participation in policy-making processes in countries where AI laws are being drafted or passed ². This has implications for the advancement of innovation and erodes the feeble relationship and trust between people, industry and technology.

The EU AI Act and The Brazil Artificial Intelligence Bill

The EU AI Act draft proposal and the Brazil Artificial Intelligence Bill are commendable initiatives toward a safe global AI ecosystem and address some of the challenges similar to those encountered in the Wisconsin v Loomis case. The two legislative instruments take different perspectives and motivations to govern AI. While the EU aims to become a global leader in artificial intelligence based on “EU norms”, a yet undefined term, the Brazilian government’s objective is to counter the impacts of foreign-produced AI on its society. This chapter is not a criticism of these two initiatives but a conversation pointing out the pitfalls that can be fixed to make certain a sustainable, innovative, ethical and moral outcome built on principles of justice, equality, inclusion and the rule of law. Upholding the above values will ensure that the AI environment is conducive to advancing innovation and creativity while maintaining respect for human dignity. It also advocates for universal governance principles or some sort of global benchmark standards by shunning nationalism and protectionism in AI policies which might hinder the use of globally produced AI interventions where they are needed most. This universal approach considers that technology has no borders and will impact societies within and beyond national borders. But let’s talk about why the status quo must give.

Just like its Brazilian counterpart, the AI Act draft lacks the benefit of substantive input from diverse social groups that could have enhanced its quality. On the 20th of February, 2020, something remarkable happened in the history of AI in the EU region—a window to a democratic process was opened, and it closed on the 14th of June 2020. You might not have heard about this process, and you aren’t alone. Or, perhaps you did and even participated. If so, you are one of the only thousand-plus out of over four hundred million EU residents and citizens who contributed to shaping the future of AI ³. To those that didn’t hear about it, a public consultation round was opened to supposedly allow stakeholders from diverse backgrounds and affiliations to share their views about the proposed policy options on AI. The list of the target population was short of being meaningfully inclusive. Though it included civil society organisations and citizens among its seven participation categories, the participation levels, as indicated in the proportion of participants per category, is discouraging. Only 1216 valid responses were recorded through online surveys, which were available on the dedicated EU Commission website for four months. The question that remains is whether the means for soliciting public opinion as a democratic process in public policy-making are adequate for AI governance. Or should the voters be allowed to decide on the proposed AI policy options?

Learning from history for a resilient future

History is replete with practical lessons in human governance failure that threaten the world with multiple-systems failures in areas where the adoption of AI systems would have a different outcome. Since 2020 and counting, the COVID-19 pandemic has inundated global and national health systems. Before its recovery, another pandemic, “monkeypox”, is rising. You’d be forgiven for expecting those in the top echelons of power to have learnt something from these two pandemics. Alas, it’s business as usual. No concrete actions are being taken to prepare for future pandemics. Instead of leveraging AI and other digital technologies to develop resilient pandemic-proof global health systems, selfish politics rather than global public interests are dictating COVID-19 response strategies. The overturning of Roe v. Wade in the US further weakens health systems by upending protections and rights for women’s access to safe abortion care and services. This case doesn’t only impact women in the US but has a far-reaching global effect owing to the role the US plays in international and global health politics.

The issues above are broader and more complex than they are simplified. Still, they make bare the systematic failures attributed to mistakes in traditional ways of doing politics both at local, national, regional and global levels. Some of these failures are attributed to the inability of natural human intelligence to process vast amounts of information in a limited time and to connect the dots in the information to make informed and rational decisions. As a result, information that doesn’t conform with what the decision-maker already knows or anything that contradicts their interests in the matter is ignored. The successes of AI compared to human-only decision-making processes amplify the case for the rapid adoption of AI to support or replace human intelligence where it is failing. I’m not attempting to present AI systems as a panacea for social injustices and all human problems⁴. Still, AI systems like COMPAS provide more logical and, to some extent, objective solutions than their human counterparts.

Unlike human-only decision-making, AI decision-making processes can be systematically audited and rectified using various tools, some of which are open source⁵. Robustness, explainability and fairness, among other tests, increase the transparency of AI and help to account for correlations of variables that influence certain outcomes and anomalies in algorithms. In contrast, it remains a mystery to understand what motivates judges, politicians and bureaucrats to arrive at certain decisions. It becomes incredibly challenging to predict the probability of a particular outcome in human-only decision-making processes. The consequence is a lack of planning, which also impacts future societies' sustainability. Despite AI’s predictability and low error margins, there’s no consensus on what constitutes its moral and ethical principles. Of particular concern, the Wisconsin v Loomis ruling and the definition of transparency, fairness and bias by the judges call into question the need for universal baseline definitions, especially for global AI. The same also goes for practically applying these ethical principles in real-time.

Policy-making and the governance ecosystem

The lack of AI regulations in many countries and the global ecosystem presents challenges for the acceptability of AI in societies and has negative implications, especially for vulnerable social groups. This creates a vacuum in governance where courts have often intervened with rulings that lead to systematic discrimination against ethnic minorities and other protected social groups in many countries. The land rights disputes between the Botswana government and the BaSarwa ethnic minority living in the Central Kalahari Game Reserve is one example that warns of the consequences of the lack of AI legislation and regulations. While other ethnic groups have entitlements to tribal lands, there are no laws recognising the BaSarwa’s rights to their tribal territories. In the absence of these constitutional protections, the judiciary has tossed them from one court to another and often ruled against them, stripping off their native and citizen rights.

The above case highlights the need to build resilient governance ⁶ systems that respect human dignity while limiting subjectivity, irrationality, and human error  in governance. To limit bias, subjectivity and human error, AI policies must consider local people and their relationship to their local environments and resources; however, without treating local issues as existing in isolation from the outside world but as situated in a relationship and communicating with each other. By adopting AI in social justice, governance systems will ensure that all people have equal access to basic human needs and that every individual and community has easy access to their share of natural and national resources, not as a privilege but as an unalienable right. How do we then ensure that the AI algorithms aren’t going to replicate the very same problems of bias and inequality?

Challenges of governing a globalising world

To some marginal extent, technology has succeeded in gradually integrating the world into one big global village where essentially anything can be conducted without needing to be in one geographic location. Subsequently, human behaviour, attitudes and social relationships are transformed to shape new human experiences. The present experiences with AI have also shown us the work that still needs to be done in person, which AI cannot do in its current state. Accelerated by the COVID-19 pandemic, the technology demonstrates its potential to bring the world together in real-time as lectures, conferences, and family meetings, among other social events, are conducted virtually for over two years. Due to restrictions, technological companies made profits from the global connectedness and use of digital technologies without in-real-life meetings. Some universities had record-high enrolment levels due to the convenience of digital technologies provided to learners and educators. Students completed their studies without meeting on campuses and attending conventional classrooms.

The COVID-19 lockdown also showed the negative side of over-dependence on technology. The lack of in-real-life meetings caused mental health problems and hindered students from creating networks crucial in adult life. A cheaper alternative replaced the privileges of those who had the right to travel, and above all, some organisations suffered because not all tasks could be conducted remotely. The gist of this argument is that technology permeated geographic and political jurisdictions to impact the lives of people who might not have access to those services due to states’ immigration, education and other national policies. Companies and organisations seamlessly worked with their staff from all over the globe without worrying about rigid immigration policies. Despite these undeniable benefits, AI shouldn’t spell the end of in-person interactions and shouldn’t be presented as the panacea for all human problems but as an alternative system that augments rather than replaces human intelligence, labour and traditional social processes. As the world continues to connect across borders and boundaries, the need for universal rules governing universal spaces increases, lest anarchy and despondency disrupt technological enthusiasm.

Alternative means of doing politics

AI is providing humanity with the opportunity, the cause and the necessary tools to imagine a new means of governance and political organisation. This depends on the people’s will to move away from the old ways of doing politics. We have the opportunity to decide how we intend to shape the rules that will impact and form the dreams we conceive for sustainable AI imaginaries. AI also provides practical tools to shape the new era of governance. In its infancy, it offers a valuable and unique platform for substantive political participation and communication where public opinion can be created, tracked and recorded in real-time, allowing for the development of models that can predict how societies will behave in the future. These platforms will also enable the propagation of a healthy democratic system where diversity and tolerance thrive through encouraging open discourses and informed alternative preferences that respect the rights of others. Rather than reacting to social trends and playing catch with policy-making, AI will allow prospective decision-making. Thus, political and business decision-making aligns with social changes in real-time.

However, if rule-making is the responsibility of those elected by people of specific territories or countries, who should elect the representatives to make the laws governing global AI? Suggesting a universal governance framework or a cosmopolitan democracy is unimaginable given the practical examples drawn from international governing bodies such as the UN, WHO and the ILO, among others. It takes generations to agree on the most basic and common-sense issues, such as climate change and universal health coverage, let alone AI. On the other hand, if individual countries develop their regulatory frameworks, the risk is that political interests will be put ahead of public good, consequently slowing technological innovations from reaching places they might impact. Through a unified or fragmented approach, citizens must decide which form or forms of governance they prefer.

Nevertheless, an inclusive global government that moves away from national citizenship is desirable. It will ensure that AI and governance adhere to local and international standards while universal democratic principles guide governance processes at all social levels. In the event of a universal AI governance, inclusion will also ensure that the future is an outcome of the processes that include all social groups impacted by AI interventions.

Governing with the periphery and the margins

While various democratic processes are rolled out as part of AI policy-making worldwide, they are inaccessible to society's marginalised and peripheral members⁷. For example, many interest groups have aired their concerns regarding how Brazil’s House of Representatives passed the Brazil Artificial Intelligence Bill without exhausting public consultation processes. Another problem is that languages and terminologies used in draft AI policy documents are too advanced and detached from the general population’s comprehension. In post-colonial states, colonial languages are still used as official languages, which only a few can fluently converse. In some countries, the venues where public consultations are held are inaccessible to those on the periphery of societies.

Adding to the exclusion list, consultations are held at inconvenient times when most people are at work, school or other activities. As a result, the working class must choose between working to keep their jobs or attending meetings. In some cases, meetings and discussions surrounding policy-making are high-level such that participation is only by invitation. As a result, the most significant demographic chunk of society is often excluded from policy discussions. Their attendees are usually from the upper classes of society—the rich, the learned, the affluent and those with a college education. Even when attempts are made to address representation, participation and inclusion, the efforts are just window dressings. Ordinary people’s views are often not reflected in the final products or are mentioned in passing, rendering traditional public engagements ineffective and wasting people’s time. These issues are still present in the current process surrounding AI legislation-making processes. Hence, the need to reconsider how we think about politics when shaping AI regulation if we are to facilitate a sustainable and inclusive transition into the future.

As I sum up, one thing becomes apparent: the routes taken to shape the future of AI are an unstable foundation to build on sustainable future imaginaries. We can’t afford to have a future that is a replica of its predecessor, which has threatened the world with multiple systems failures. However, that doesn’t mean we need to reinvent the wheel. What is needed is a gradual, step-by-step and systematised departure from traditional governance while determinedly hastening the transformation of political organisation. Also, AI’s future shouldn’t be treated as a given or a process unfolding from thin air but as a culmination of change processes that begin now and are taken by everyone for everyone. New approaches to AI governance should also rectify that participation in a democratic process isn’t a privilege offered to the public but a right that every citizen and resident of a given community, country or any political jurisdiction is entitled to. In that rectification, the burden of inclusion shouldn’t only lie with institutions and governments alone. Citizens must be willing to move from consumers of rules made in black boxes of governance to active producers and participants in shaping AI norms at home and globally.