In a War of Imbalance, Quantity Doesn't Equal Security

A completely unregulated arms race in artificial intelligence (AI) does not lead to effective solutions, despite the belief that more accessibility results in more good AIs than bad. The defense and attack dynamics are asymmetrical; attackers exploit vulnerabilities at a faster rate than defenders can patch them. For instance, in cyber warfare, an attacker can succeed if they find a single weakness, whereas defenders must close every possible gap. Similarly, if AI is weaponized to develop bioweapons, attackers can take their time to create and deploy lethal weapons while defenders may require significant time to identify and cure outbreaks. Therefore, mere numerical superiority of good AI controllers does not guarantee safety; the complexities of the landscape require robust strategies rather than a loose distribution of power.