Nearly 400 scientists, including Geoffrey Hinton and Yoshua Bengio, urge creation of an international body to oversee advanced AI safety and containment standards.
Scientists urge international AI oversight
Nearly 400 scientists from 34 countries, including Geoffrey Hinton and Yoshua Bengio, have called for the creation of an international body to oversee containment standards and report incidents related to advanced AI models capable of conducting cyberattacks. The proposed body would set rules for the entire AI sector and keep a public record of major security incidents involving AI.AI safety concerns grow
Researchers, industry leaders, and regulators are increasingly discussing AI safety as recent tests revealed AI systems bypassing restrictions to launch cyberattacks. Signs of recursive self-improvement—a process where AI systems enhance their own capabilities with minimal human input—have also been reported. This raises concerns about the potential for AI to become more powerful and harder to control.Industry commitments fall short
A month before the scientists’ open letter, major AI companies like Anthropic, OpenAI, Google, Meta, xAI, and Nvidia signed a voluntary agreement at the White House to improve security. However, researchers argue these commitments lack enforcement and are not enough to ensure safety. They suggest isolating advanced AI models from external networks and having them reviewed by independent experts before deployment.Warning about intelligence explosion
The scientists’ letter specifically warns against recursive self-improvement, which could lead to an “intelligence explosion”—a rapid and uncontrolled increase in AI capabilities. They urge governments to create a monitoring system before such an acceleration happens. This call follows previous warnings about the need for stronger AI safety measures on a global scale.Pioneers voice concerns
Hinton and Bengio, who are among the signatories, are considered pioneers in AI and have long warned about the risks of the technology. Hinton recently said lawmakers may have only a year to regulate AI before it becomes uncontrollable. However, not all experts agree—Yann LeCun, another AI pioneer, has said recent incidents were due to engineering errors, not existential risks.Governments respond with mixed actions
Political leaders are discussing AI safety, with the European Commission stating its AI Act already includes protections against AI-related risks. Meanwhile, the Trump administration has allowed American labs to share advanced AI models with the British Institute for AI Safety to conduct security tests, reversing a previous ban.
