Experts from the United States and China in national security and artificial intelligence have proposed a new safety framework for advanced AI systems. The initiative suggests adopting risk management strategies similar to those used for nuclear weapons during the Cold War to mitigate the risks of technical failure, miscalculation, and escalating international competition.
The report highlights that future AI systems could play critical roles in cybersecurity, military command, intelligence analysis, and infrastructure management. The authors warn that unexpected behaviors or malicious use of these technologies could have global consequences, necessitating a multi-layered security approach including improved communication channels and risk assessment cooperation.
A key recommendation involves creating a 'hotline' mechanism to ensure immediate contact between nations during major AI-related incidents. This is intended to prevent misjudgments caused by mutual distrust, which the experts identify as a primary danger. The proposal also calls for stricter safety reviews, independent testing, and continuous monitoring for the most advanced AI models.
While acknowledging that AI and nuclear weapons are distinct, the researchers emphasize that both technologies carry significant strategic weight. They argue that establishing protective frameworks early is more effective than reacting after risks materialize. The experts suggest starting with pragmatic risk management rather than seeking comprehensive regulatory agreements immediately.
The report notes that current global AI governance remains in its early stages with significant gaps in cross-border collaboration. Given the rapid pace of technological development, the experts stress that the window for international cooperation is limited. They conclude that the US and China bear a special responsibility to reach a consensus on safety standards and crisis communication.
