Altman Warns UN on AI Risks, Calls for Global Safety Pact

Overview

OpenAI CEO Sam Altman addressed the United Nations Security Council on Wednesday, delivering a stark warning about the existential risks posed by rapidly advancing artificial intelligence. In a speech that emphasized both the transformative potential and the immediate dangers of the technology, Altman called for a unified international approach to safety governance.

He argued that humanity stands at a critical juncture where AI could either spark a new Renaissance of creativity and discovery or trigger an era of unprecedented upheaval. The CEO stressed that the pace of progress has compressed timelines, making the stakes more tangible and the risks more urgent than ever before.

"We have a choice in front of us. AI can either be more like a new Renaissance of creativity and discovery, or more like a new Industrial Revolution of upheaval and disarray."

Altman’s remarks came amid growing global anxiety over recursive self-improvement and the concentration of power within a few powerful entities. He urged member states to reject the notion that competitive advantage justifies reckless deployment, advocating instead for rigorous safety guarantees and human oversight.

The Dual Nature of AI Progress

The OpenAI chief framed the current moment as one defined by dual potentials. On one hand, he highlighted the technology’s capacity to solve humanity’s most pressing challenges, from discovering new medicines and improving healthcare to strengthening education systems and boosting economic productivity. He described AI as a defensive tool against the increasing complexity of modern life, empowering individuals to pursue ambitions that previously seemed out of reach.

On the other hand, Altman cautioned that as systems become more capable and autonomous, they risk outpacing institutional safeguards. He warned that these systems could concentrate power in too few hands, potentially allowing a single person, company, or nation to impose its worldview on the rest of the world. This dynamic, he argued, undermines democratic principles and threatens global stability.

Altman rejected the dystopian narratives often circulated in popular culture, asserting that the goal of AI should not be to optimize life until human agency disappears. Instead, he envisioned a future where technology expands human capability, allowing people to learn, create, build, and participate more effectively in an increasingly complex world.

Risks and the Case for Control

A central theme of the address was the danger of losing control over AI systems, particularly as they approach the threshold of recursive self-improvement. Altman noted that as the process of building AI becomes more automated, the pace of progress could accelerate rapidly, potentially leaving human institutions unable to intervene when necessary.

He explicitly aligned himself with the concerns raised by prominent researchers such as Yoshua Bengio, identifying two primary pathways for catastrophic failure. The first is the loss of human control over the future due to the speed and opacity of AI development. The second is the excessive concentration of power, where a single actor monopolizes the most advanced models.

"None of these levels are remotely acceptable. And we should not train models that we cannot make an extremely strong case that we will be able to keep under human control."

Altman emphasized that no percentage of risk—whether estimated at 1%, 10%, or lower—should be considered acceptable if it involves deploying systems that cannot be reliably monitored or controlled. He argued that the industry has a moral obligation to avoid building systems where alignment and safety guarantees cannot be firmly established, regardless of the competitive pressure to move faster.

Principles for a Global Safety Pact

To navigate these challenges, Altman outlined three foundational principles that he believes should guide the development and deployment of artificial intelligence globally.

First, he asserted that as AI becomes more capable, people must remain at the center of decision-making. He defined alignment not as an abstract research problem, but as the practical work of ensuring systems reliably reflect human values and remain under human direction. Automation of human judgment, he insisted, is not the goal; rather, the focus must be on enhancing human agency.

Second, the CEO argued that no single entity should possess the unchecked ability to deploy powerful AI without accountability. He warned against the logic that trustworthiness justifies authoritarian control, stating that any company or country claiming exclusive right to such technology is laying the groundwork for dangerous overreach.

Third, Altman called for international cooperation to establish safety standards that transcend national borders. He suggested that OpenAI itself has a role to play in modeling responsible behavior, citing instances where the company has unilaterally slowed down development to prioritize safety. He affirmed that this approach would continue, rejecting the idea that the industry is locked in a race that prevents it from exercising caution.

"Beating companies in a competitive pace is not a reason to make rash decisions. Nor do we believe we are locked in a race where we are unable to do that."

Conclusion

Altman concluded by urging the Security Council to walk a middle path between blind optimism and fatalistic doomerism. He argued that neither extreme helps solve the enormous challenges facing the world today, many of which AI could help address if managed responsibly.

The CEO’s speech served as a direct appeal to world leaders to treat AI safety as a matter of global security, akin to nuclear non-proliferation. By calling for a global safety pact, Altman aimed to shift the discourse from competition to cooperation, ensuring that the benefits of artificial intelligence are shared broadly while the risks are mitigated collectively.