Vivian Balakrishnan on AI Safety at UNGA

Caption language: en · Fetched: 2026-09-28

Do you want the nuclear button to be in the hands of a human being or an AI? Think about that. AI offers what extraordinary potential, but its capability is advancing faster than our appreciation of its risks. But excellencies, the genie is already out of the bottle. It is too late to call for a halt, especially given the superpower rivalry and the huge financial incentives for companies who already have frontier models. But let me offer a metaphor of driving a high-performance car. To drive a high-performance car quickly and safely, you need a powerful engine. But equally, you need good brakes. And especially if you're going to go beyond your neighborhood or cross boundaries, you need effective shared rules, traffic lights, insurance, norms of behavior, seatbelts. The engine is innovation.

But having said that, every powerful engine needs good brakes, safeguards. As the capabilities of AI advance, our safeguards must keep pace. We need rigorous testing and evaluation before deployment. We need clear limits on what autonomous systems can do and mechanisms to intervene when they behave in ways we do not intend. And since the risks from AI do not remain confined behind borders, it means we also need rules of the road that will form a shared global understanding of how we will collectively confront this challenge. Just as all of us instinctively understand traffic lights the same way, whether we're here in New York or in Beijing, Moscow, or Singapore. Most importantly, human beings must ultimately remain in control and accountable. Let me give you an extreme thought experiment.

Do you want the nuclear button to be in the hands of a human being or an AI? Think about that. The scarce resource today is not legal architecture or technology. It is trust. Trust that the risk will be disclosed, that testing will be comprehensive and credible, and that cooperation will be inclusive and not be used to secure unilateral advantage.