At the 2025 United Nations General Assembly, a powerful group of more than 200 global leaders—among them Nobel laureates, renowned scientists, former heads of state, and senior AI developers—has urged governments to establish clear international “red lines” for artificial intelligence before the end of 2026. 

Growing concerns over AI risks 

The statement highlights how advanced AI systems are already exhibiting troubling patterns, such as deceptive behavior, autonomy in decision-making, and the ability to influence large-scale outcomes without sufficient human oversight. Experts warn that these unchecked capabilities could lead to catastrophic consequences, including: 

• AI-enabled biological threats and engineered pandemics 

• Widespread job displacement caused by automation 

• Loss of meaningful human control over powerful systems 

The group argues that unless global boundaries are drawn soon, the risks of AI could quickly overshadow its potential to transform healthcare, education, and economic development. 

A diverse coalition of voices 

What makes this call significant is the breadth of its supporters. Notable signatories include AI pioneers Geoffrey Hinton and Yoshua Bengio, Nobel Prize-winning economist Joseph Stiglitz, and former political leaders Mary Robinson and Juan Manuel Santos. 

The statement has also drawn support from researchers at leading AI labs such as OpenAI, Google DeepMind, and Anthropic. However, the CEOs of these companies have notably not signed on, underscoring the divide between developers and corporate leadership. 

What do “red lines” mean? 

Although the proposal avoids detailing exact restrictions, it builds on earlier ideas from the AI safety community. Suggested prohibitions include: 

• Autonomous replication of AI systems 

• Development of destructive or uncontrollable power-seeking behaviors 

• Large-scale cyber operations conducted without human supervision 

The signatories emphasize that these boundaries must be not only defined but also operational and enforceable by 2026, with international mechanisms to ensure compliance. 

Challenges to enforcement 

Creating enforceable AI boundaries will not be easy. Different nations have competing priorities—while some call for strict global regulation, others, such as the United States, remain wary of sweeping restrictions that could impact innovation or national competitiveness. 

Another obstacle is agreement on definitions: what qualifies as “unacceptable risk” can vary widely across governments and industries. Effective red lines would require unprecedented international cooperation, robust enforcement structures, and transparent monitoring mechanisms. 

Why the call matters 

The demand for red lines reflects the urgency of the present moment. Humanity is entering an era where artificial intelligence may rival or exceed human intelligence across multiple domains. Whether these technologies drive progress or generate instability depends heavily on the decisions made today. 

By placing AI governance at the center of global dialogue, Nobel laureates and AI researchers are pressing world leaders to act before it’s too late. Their message is clear: responsibility must match capability, and the future of AI should be guided by safety, accountability, and collective human values.