By Irakli Beridze
Head, UNICRI Centre for Artificial Intelligence and Robotics
Imagine a world in which scientific discoveries that once took years happen in weeks, doctors detect diseases earlier, and public services become more accessible and effective. This is no longer simply a vision of the distant future. Artificial intelligence is already bringing parts of it within reach. Google DeepMind’s AlphaFold, for example, has made predictions of more than 200 million protein structures available to researchers worldwide, opening remarkable possibilities for scientific discovery.
As AI systems become increasingly capable, their potential to transform our societies will only grow. There is every reason to be optimistic. Yet, this positive potential rests on a question that receives far less attention than technological breakthroughs themselves: are our institutions prepared to make the most of it?
Since establishing the UNICRI Centre for Artificial Intelligence and Robotics in The Hague in 2017, we have seen how the difference between technological potential and meaningful impact often lies in institutional capacity. Our AI for Safer Children initiative, developed with the United Arab Emirates Ministry of Interior, demonstrates how AI can deliver tangible benefits. Supporting investigators across 131 countries in combating online child sexual exploitation and abuse, the initiative illustrates what becomes possible when innovation is combined with skilled professionals, international cooperation and institutional responsibility.
This experience also underpins AI-POL, our inititative with INTERPOL, supported by the European Union, to help law enforcement institutions adopt AI responsibly. Effective adoption requires attention to the entire lifecycle, from understanding technological capabilities and limitations to development, procurement, deployment and continuous oversight. Leadership, accountability and public trust are equally essential. But institutional readiness cannot be measured solely by an organization’s ability to adopt AI. It must also be measured by its ability to understand, question and, when necessary, reject the conclusions these systems produce.
The same challenge extends beyond law enforcement. Governments, diplomatic services, judicial institutions and international organizations are increasingly integrating AI into their operations. Their ability to do so responsibly depends not only on technological expertise, but also on preserving the institutional judgement and accountability needed to govern its use.
Having participated in international AI governance discussions for more than a decade, I have witnessed considerable progress in developing shared principles, frameworks and mechanisms for cooperation. These achievements are important; but international consensus, however valuable, is not effective without institutional capacity. The real test of governance is not only what governments agree upon, but whether their institutions can translate those commitments into responsible decisions and everyday practice. International cooperation must therefore move beyond setting expectations to helping institutions develop the capacity to meet them.
This makes AI readiness a strategic priority for institutional leadership, rather than simply a technical or compliance responsibility. Governments and international organizations must invest in the people, decision-making processes and organizational cultures that allow AI systems to be evaluated and challenged. Organizational culture is key, since even institutions with strong governance frameworks face the more subtle challenge: maintaining human oversight on paper does not necessarily mean preserving it in practice.
As AI becomes more capable, we will naturally delegate more tasks to these systems. Much of this will be beneficial, expanding human capabilities and allowing people to focus on more meaningful work. But we must distinguish between delegating tasks and surrendering judgement. When sophisticated recommendations arrive instantly and with apparent confidence, the temptation to accept rather than question them grows. Over time, institutions may retain formal responsibility while gradually losing the practical capacity to independently verify, explain or challenge their decisions. For public institutions, where decisions can affect fundamental rights, justice and public trust, this responsibility is especially consequential.
The greatest danger may arise when increasingly capable AI systems meet institutions that have gradually surrendered their capacity for independent judgement.
This is not an argument against technological progress. Quite the opposite. It is an argument for ensuring that our institutions evolve alongside it – becoming not only more technologically capable, but also more resilient, accountable and confident in exercising human judgement with the support of AI.
I remain profoundly optimistic about what AI can help humanity achieve. Its future is not predetermined by technology alone. It will also be shaped by the institutions we build, the values we uphold and the decisions we make.
The choice is still ours. We must remain the authors of our history, not the characters of a story we did not write.
