Irakli Beridze – Can the UN Help with Global AGI Governance? (AGI Governance, Episode 11)

ByAnnelize P October 10, 2025October 10, 2025 October 10, 2025

Joining us in our eleventh episode of our series AGI Governance on The Trajectory is Irakli Beridze, Director of the UNICRI Centre for Artificial Intelligence and Robotics under the United Nations mandate. His remit spans AI’s intersections with crime prevention, criminal justice, rule of law, and human rights – and he’s spent years convening governments, industry, and researchers on AI’s risks and opportunities.

In this conversation, Irakli draws a stark contrast between yesterday’s arms-control templates and tomorrow’s AI. Chemical weapons were narrow, outdated, and ultimately unwanted. By contrast, advanced AI is general-purpose, fast-accelerating, and universally desirable – which makes governance both harder and more urgent.

He also underscores inclusion as a first principle. If the technology will affect everyone, then everyone should have a voice – and some stake – in how it is governed and how its benefits are distributed. Waiting for a forcing crisis, he argues, risks arriving after control is already lost.

I hope you enjoy this conversation with Irakli:

Subscribe for the latest episodes of The Trajectory:

Below, I’ll summarize Irakli’s main points from each of the four sections of our interview.

AGI Governance Q-and-A Summary – Irakli Beridze

1. How important is AGI governance now on a 1-10 scale?

10 out of 10. 

Irakli’s position is unambiguous: the time to build the foundations for governance is not when AGI arrives – it’s now. He stresses that establishing international frameworks, institutions, and norms takes years, even decades. The chemical and nuclear treaties that many point to as precedents each required long periods of negotiation, iteration, and verification before they became functional. Waiting until AGI systems surpass human control, he warns, would mean trying to retrofit governance mechanisms around entities that already outthink us.

For Irakli, early coordination isn’t optional – it’s the only way to ensure humanity retains agency as capabilities grow. He argues that governments should already be building the scaffolding of a future system: creating shared definitions, raising awareness, and mapping how verification might actually work. 

2. What should AGI governance attempt to do?

Irakli believes AGI governance must begin by bringing the topic into formal international discussion. He explains that neither his center nor the UN at large are yet working on AGI-specific structures, and that recent UN General Assembly debates and resolutions have not included it. For him, this absence shows how far institutions still are from preparing for the systems now in development, and why initiating these conversations early is essential.

He also stresses that governance must address both sides of the equation – managing risk and distributing benefits. Irakli warns that if AGI development and control are limited to a small number of actors, it could create deep inequalities and instability across nations.

Finally, he emphasizes urgency. The acceleration of AI capabilities, he says, means that the world cannot afford to wait. Governance discussions and frameworks must start before AGI arrives, because by the time it does, it may already be too late to shape safely.

3. What might AGI governance look like in practice?

Irakli envisions a model built on international agreement and verifiable compliance. He explains that countries will eventually need to reach consensus on parameters for “powerful AI” – including rules to ensure systems remain controllable and do not cause harm. For that consensus to work, he says, it must be paired with mechanisms that prove adherence rather than rely on diplomatic trust. The foundation of governance, in his view, is trust through verification.

He also points to the possibility of using hardware-based monitoring tools as one piece of the puzzle. Irakli supports the concept in principle but stresses that no single technological approach can stand alone. He suggests that hardware checks may complement – not replace – other verification systems and that governance should explore several methods in parallel.

4. What should innovators and regulators do now?

For innovators,Irakli says innovators have a direct responsibility to raise awareness of AGI among policymakers. He stresses that scientists, engineers, and industry leaders must bring these conversations to the highest levels of government – including heads of state – because many still have no understanding of what is coming. Innovators, he explains, should help delegations and decision-makers grasp the trajectory of AI development and the pace at which capabilities are growing. They should show, through demonstrations and examples, that systems far more powerful than those we have today are close at hand. He adds that private-sector participation will be indispensable in any treaty or global governance process, just as it was in the chemical-weapons framework. Above all, he urges innovators to begin building governance options now and to make AGI a mainstream topic before the world is overtaken by events.

For regulators, Irakli argues that regulators and governments must move AGI governance from concept to formal policy. He notes that the UN has yet to mention AGI in any General Assembly resolution, and that this silence must end. He insists that any framework must be genuinely global: a technology that affects everyone cannot be governed by only a few states. Regulators, he says, should begin defining parameters for “powerful AI,” create verification mechanisms to ensure compliance, and establish enforcement powers that do not depend solely on interstate trust. He warns that if governments wait until AGI systems are deployed, control could already be lost – because once such systems exist, they will evolve faster than human oversight.

Irakli’s message is clear and urgent: artificial general intelligence will be wanted everywhere, embedded in everything, and advancing faster than most institutions can adapt. Governance, he argues, must begin before that reality arrives in full – grounded in broad participation, credible verification, and open political acknowledgment at the highest levels. Waiting until later, he warns, would mean reacting to a force already beyond control.

Yet he ends on a note of cautious optimism. For Irakli, humanity’s defining strength has always been its ability to recognize danger and unite around solutions when it matters most. He believes that same instinct can prevail again – if coordination starts now.

…

Follow The Trajectory

• Sign Up for The Trajectory Newsletter
• Subscribe to The Trajectory on YouTube
• Subscribe to The Trajectory podcast
• Follow on LinkedIn
• Follow on X