F
rom ChatGPT and Grok to autonomous drones, facial recognition, and predictive policing, Artificial Intelligence (AI) is changing the way we live, work, and govern. However, our ability to control the political, moral, and ethical hazards of AI is outdated. National laws regarding sophisticated technology are limited and mostly outdated in the age of AI. Meanwhile, AI systems and technologies are being used to target marginalized populations, displace labor, influence elections, spread misinformation, and enable widespread surveillance of society.
The rapid advancement and applications of AI have outpaced our capacity to regulate it. In order to ensure that this technology benefits society, we urgently need a global AI governance framework — a set of internationally agreed-upon laws and guidelines designed to mitigate the risks of AI.
Currently, there is no global framework to enforce accountability, no uniform mandate to ensure algorithmic transparency and equity, and no international legislation to prevent the deployment of dangerous AI systems worldwide. In this regulatory vacuum, global tech firms and dictatorial regimes are vying for control of AI, sometimes with little consideration for morality, privacy, or human rights. This new reality is what American philosopher Shoshana Zuboff terms “surveillance capitalism.”
I argue that this issue could be resolved through a global framework establishing uniform safety, transparency, and equity criteria while preserving national autonomy, security, and the ability to customize policies according to unique circumstances. This framework would strike a balance between national sovereignty and community responsibility, similar to the Paris Climate Agreement or international nuclear weapons treaties.
First, AI must be grounded in human rights, dignity, and human-centered universal norms. AI systems should respect non-discrimination, freedom of speech, privacy, and individual and communal dignity. AI systems, particularly those trained on biased data, often perpetuate injustice, racism, and inequity. Facial recognition technology is used in some countries to track and silence political opponents. Without rights-based governance and regulations, AI has the potential to become an instrument of tyranny.
Second, AI systems must be understandable and transparent. If someone is targeted by the police, denied a loan, or hired or fired based on an algorithm, the reasoning behind the decision must be made clear. Black-box systems undermine public confidence and make it difficult to contest incorrect results. AI technology is neither objective nor impartial; it is mostly trained on biased, Eurocentric, and White-centric data. For example, AI systems used by police in the United States disproportionately profile Black people, thus perpetuating racism. Technology is not neutral; it reflects the political and moral sensibilities of its creators and programmers.
Third, the entire AI lifecycle must incorporate accountability. In order to audit systems, investigate complaints, and impose sanctions, independent regulators must have the authority to analyze and regulate AI systems. Victims of algorithmic damage must have channels to receive compensation in cases of AI discrimination, rights abuse, and surveillance. A few weeks ago, Grok wrote insults on X. A global AI governance framework is essentially corporate self-regulation in the absence of national laws.
Fourth, AI systems must be trustworthy and secure. Failure of an AI system could have disastrous consequences in high-stakes fields such as healthcare, transportation, education, and military and security applications. To mitigate global catastrophes, continuous monitoring, impact analyses, and rigorous testing are non-negotiable.
Fifth, AI development must be inclusive, democratic, and equitable. Currently, the majority of the world’s AI research and infrastructure is controlled by a few tech companies primarily located in the Global North. While businesses in Africa and other developing nations are disrupted by AI technologies and their data continues to be collected en masse, communities in the Global South are often excluded from designing AI systems and processes. A global framework that challenges digital capitalism, the digital divide, and inequality must guarantee the capacity-building and fair participation of poor and developing countries.
Sixth, we must address the environmental consequences of AI. The amount of energy required to train large language models is staggering. As the climate crisis worsens, the AI industry must support and fund larger environmental objectives, not undermine them, in order to have sustainable, green AI.
In short, international collaboration is crucial. Disinformation, cyberwarfare, and autonomous weaponry are a few of the many worldwide risks associated with AI. No nation can handle these risks alone. An urgent need exists for a global policy framework that safeguards humanistic ideals, encourages collaboration, and lessens the possibility of AI-driven weapons races.
Some groundwork is already being established, which is encouraging. The EU just enacted the world’s first comprehensive AI law. In 2021, UNESCO published global AI ethics guidelines, and in 2023, 28 countries signed the Bletchley Declaration, committing to the safe and sustainable development of AI. However, these are only preliminary measures. What is lacking is a legally binding, enforceable global agreement addressing the full range of AI risks and responsibilities, potentially under the United Nations’ auspices.
Some might argue that such cooperation is impossible due to geopolitical tensions, particularly between the U.S. and China. However, the same was true throughout the Cold War with nuclear armament accords. Nevertheless, imperfect collaboration is preferable to none on global existential matters like AI. If we fail to act now, we risk handing over too much power to machines we don’t fully understand — as well as to the unaccountable individuals who control them.
Stay informed by subscribing to
our weekly newsletter.
