Sam Altman, CEO of OpenAI, delivered a speech at the United Nations Security Council on AI issues, discussing the opportunities and risks brought by AI development and calling for enhanced international cooperation. He advocated for establishing standards for capability assessment, risk assessment, safety measures, and human oversight for cutting-edge AI systems. He stated that as AI capabilities rapidly improve, the potential benefits and risks of the technology have become more directly evident—AI could be more like a renaissance full of creativity and discovery, or it could resemble an industrial revolution causing turmoil and chaos.
Two Main Lines: Risk of Losing Control and Concentration of Power
Altman summarized the serious issues that AI might bring into two directions. First, losing control over the future: AI is progressing too fast, people cannot keep up, nor can they intervene when needed. He clearly stated that the industry should not accept excessive technological risks just because AI is so profitable, nor should it make hasty decisions due to corporate competition; "OpenAI has slowed down unilaterally in the past and will do so again in the future." It should not train models that are unable to be controlled by humans, regardless of whether the disaster risk is estimated at 10%, 1%, or 0.1%, all of which are unacceptable. He specifically mentioned recursive self-improvement (the process where AI participates in AI development), stating that if development becomes further automated, the pace of technological advancement may increase rapidly, requiring high caution.
The second point is the long-standing issue of concentration of power: the strongest models and related power should not be concentrated in the hands of a few people, to avoid one person, company, or country using AI to impose their worldview on everyone. Altman emphasized that this logic must be rejected, even if it benefits them. OpenAI tries to walk a middle path between blind optimism and doomsday scenarios, following three principles: humans should continue to be at the core of AI decision-making; the benefits of scientific progress and economic growth should serve the people; and technology must empower individuals, allowing more people to use the technology to achieve their own goals.
Mathematics Has Reached International Competition Level, Governance Must Follow Democratic Processes
Altman also discussed the progress of AI in mathematical research: OpenAI models could only handle elementary arithmetic three years ago, reached the level of high school competition gold medalists two years ago, and now have reached international mathematics competition levels. He gave an example, stating that a few weeks ago, an internal model solved one of the Millennium Prize Problems, the Navier-Stokes equations. However, he emphasized that solving the problem was not for the sake of mathematics itself, but to empower people to discover new knowledge and improve healthcare. Moreover, companies cannot replace democratic processes—if AI is to be "democratic," the most important decisions should not be made solely by the San Francisco lab, but must go through democratic processes shaped by governments accountable to the people, which requires cooperation at the international level.
He advocated for establishing complementary domestic and international mechanisms for advanced AI standards, measuring capabilities, assessing risks, verifying safeguards, and retaining meaningful human supervision when systems become more autonomous. It is also necessary to have accurate and rapid event reporting and secure channels for sharing vulnerabilities. At the end of his speech, Altman said the world is at a crossroads: one path leads to the concentration of power and opaque decision-making, while the other leads to AI serving the people and being determined by the people through democratic institutions, "this is the future worth building."
