OpenAI, Anthropic CEOs urge UN to adopt global AI safety framework

NEW YORK — The chief executive officers of OpenAI and Anthropic addressed the United Nations General Assembly on Wednesday, urging world leaders to establish a binding global framework for the regulation of advanced artificial intelligence. The joint appearance by Sam Altman of OpenAI and Dario Amodei of Anthropic marked a rare unified front from the industry’s leading developers, who stated that the rapid advancement of large language models has outpaced existing national regulatory mechanisms.

Call for International Coordination

Speaking at a high-level session dedicated to emerging technologies, Altman and Amodei argued that the risks associated with frontier AI models require a multilateral approach rather than fragmented national policies. They emphasized that the technology is currently at a critical juncture where international cooperation is essential to prevent misuse and ensure safety standards are met globally.

“We are at a crossroads,” Amodei said, according to reports from the event. “The capabilities we are building are powerful, and the responsibility that comes with them demands that we work together, not in isolation.” Altman echoed these sentiments, noting that while the United States and other major economies have begun to draft domestic regulations, a lack of global consensus creates vulnerabilities that could be exploited by state and non-state actors.

Context of US Policy Shifts

The CEOs’ appeal comes amid shifting policy landscapes in the United States, where recent executive actions have signaled a preference for innovation over stringent pre-market regulation. According to CNBC, the push for international cooperation follows statements from the White House that characterized certain multilateral AI governance initiatives as a “globalist scheme” to control technological development. The CEOs’ intervention at the UN appears to be a direct response to these domestic political tensions, seeking to anchor AI governance in international law rather than solely in US federal policy.

Industry analysts noted that the joint statement represents a strategic pivot for the AI sector. Historically, major AI labs have advocated for light-touch regulation to maintain competitive advantages. However, the scale of recent model releases, which have demonstrated emergent capabilities in reasoning and autonomous task execution, has prompted a reassessment of risk management strategies within the industry.

Technical and Safety Implications

The CEOs highlighted specific technical challenges, including the difficulty of verifying model alignment and the potential for catastrophic misuse in biological and cyber domains. They proposed the creation of an international body, similar to the International Atomic Energy Agency, to conduct independent audits of frontier models before their public release. This proposal aims to establish a standardized set of safety benchmarks that would be recognized by all signatory nations.

Amodei urged the industry to “slow down” in certain areas where safety evaluations are incomplete. He stated that the pace of development must be matched by the pace of safety research. Altman agreed, adding that OpenAI is currently working on new interpretability tools to better understand the internal decision-making processes of its models. Both executives acknowledged that while current models are not yet fully autonomous, the trajectory of progress suggests that such capabilities could emerge within the next decade.

International Reaction

The address drew mixed reactions from diplomatic delegations present at the UN. Representatives from the European Union welcomed the call for a global framework, noting that the EU’s AI Act, which entered into force in 2024, provides a foundational model for risk-based regulation. However, delegates from several Asian and African nations expressed concern that a US-led regulatory framework might impose standards that do not account for the developmental needs of emerging economies.

China, which has not been a consistent participant in multilateral AI governance discussions, was not represented at the specific session but has previously emphasized the importance of “responsible innovation” in its own domestic policies. The absence of a unified global stance on AI governance has been a recurring theme in UN discussions, with many countries struggling to balance economic opportunities with security risks.

The CEOs concluded their remarks by inviting governments to engage in technical dialogues with industry experts to draft a non-binding declaration on AI safety. They stated that such a declaration could serve as a precursor to more formal treaty negotiations. The UN Secretary-General’s office has indicated that it will review the proposals and may convene a dedicated working group in the coming months to explore the feasibility of an international AI safety accord.