AI shares similarities with older technologies, particularly in its "dual-use" nature, possessing applications with both military and civilian uses. Many AI systems are designed to be general, aiming for widespread application to serve as many users as possible. However, the versatility that makes them applicable to everyday tasks also means they can be repurposed for potentially destructive purposes. For instance, an AI system developed for medical diagnoses could potentially be utilized to create and weaponize new diseases.
The line between civilian and military applications of AI is inherently blurred. The same technology used by companies like Synesis to locate missing individuals in Belarus was later employed to target political activists protesting against fraudulent elections. This dual-purpose nature raises concerns, as seen in the United States' decision to restrict the export of advanced semiconductors to China.
Unlike technologies such as nuclear power, AI models are highly versatile and have the potential to be pervasive. Unlike chemical or nuclear weapons, AI systems are not prohibitively expensive to develop, store, or clandestinely share. As their benefits become increasingly evident, AI systems are poised to become more widespread, sophisticated, and affordable.
Regulating AI is a global challenge, as its proliferation is not confined to national borders. Efforts to regulate AI in one country may have limited impact if it remains unregulated elsewhere. Governing AI on an international level is crucial due to its ease of dissemination. Unlike nuclear technologies, the governance of AI requires even more extensive international cooperation, acknowledging its potential impact and ubiquity in our increasingly interconnected world.