Sam Altman Emphasizes AI Control and International Cooperation at UN Security Council

OpenAI CEO Sam Altman warned of the potential and risks of AI at the UN Security Council, urging human control over powerful systems and international cooperation. He diagnosed that AI stands at a crossroads where it could either expand creativity like the Renaissance or cause disruption like the Industrial Revolution.

Two Key Risks to AI Control

Altman presented two pathways where AI development could go wrong. The first is loss of control due to recursive self-improvement. If AI improves itself at an accelerating pace, humans may become unable to intervene. He emphasized that a risk probability of 10% or 0.1% is unacceptable, and models whose human control cannot be assured should not be trained.

The second is the excessive concentration of power. He pointed out the need to guard against situations where specific companies or nations monopolize the most powerful AI models and impose their worldview. He argued that important decisions should be made through democratic processes, not just in San Francisco labs.

Need for International Standards and Cooperation

OpenAI proposed three principles that AI should develop in a direction that empowers individuals. To this end, he suggested the establishment of Frontier AI standards at the national and international levels. Specifically, he called for the following:

• Establishment of criteria for capability measurement and risk assessment
• Ensuring accuracy and speed in incident reporting and classification protocols
• Building vulnerability sharing channels among governments, infrastructure operators, and experts

Altman added that these standards should not favor incumbent companies and must include both open and closed model developers. He concluded that AI will only benefit humanity when people proactively determine the direction of development through democratic institutions.