AI Risks Mirror Nuclear and Pandemic Threats

Recent debates have framed artificial intelligence as a potential threat to human survival, evoking scenarios where the technology operates beyond human control. While these fears may seem like science fiction, they align with historical precedents involving other existential risks, such as thermonuclear weapons and global pandemics. By examining how humanity managed these previous threats, policymakers can find a blueprint for addressing the unique dangers posed by AI.

Harlan Ullman, a former defense strategist and author, argues that the current AI landscape demands a similar level of serious oversight. He suggests that just as the world established controls for nuclear materials, a dedicated federal framework is necessary to monitor and regulate AI development. This approach would aim to harness the technology’s benefits while mitigating its potential to cause widespread harm to critical infrastructure and society.

Historical parallels guide new policy

The management of thermonuclear weapons offers a clear precedent for controlling dangerous technology. After the creation of the atomic bomb, the United States established the Atomic Energy Commission to oversee civilian use, later splitting responsibilities between the Department of Defense and the Department of Energy. This bifurcation of control helped prevent misuse and maintained stability. Ullman proposes that AI requires a similar split, with specific agencies responsible for different aspects of the technology’s development and deployment.

While climate change and pandemics also present existential risks, they differ in how they can be controlled. Pandemics, for instance, often require complex biological infrastructure that is difficult for individuals to replicate. In contrast, AI could lower the barrier for creating biological threats or disrupting power and water systems. This makes the regulatory challenge more difficult, as the technology is more accessible and fungible than nuclear materials, which are restricted to a handful of nation-states.

AI threats differ from nuclear risks

According to The Hill, the primary distinction between AI and nuclear weapons lies in accessibility. Nuclear war has been avoided largely because the technology is concentrated in state hands. AI, however, can be exploited by individuals or non-state actors to tamper with reactors or launch cyberattacks. This decentralization of risk means that traditional arms control agreements, which rely on state-to-state negotiations, may not be sufficient. A broader regulatory net is needed to catch unauthorized uses of the technology.

Furthermore, AI could amplify existing threats. It might enable the creation of highly contagious diseases or accelerate the damage to critical infrastructure. Unlike a nuclear explosion, which is a singular event, AI-driven threats could be continuous and systemic. This requires a regulatory body that is not just reactive but proactive, capable of setting guidelines and monitoring the industry for emerging dangers.

A federal agency is proposed

Ullman advocates for the creation of a specialized AI regulatory agency with the power to issue guidelines and enforce oversight. This body would work alongside the Departments of Defense, Energy, and Commerce, as well as the CIA, to identify national security threats. The goal is to establish a federal framework that checks the existential potential of AI, ensuring that the technology does not evolve beyond human oversight.

The implementation of such a framework would be a political decision, balancing innovation with safety. Internationally, the model could be replicated under the United Nations, similar to the International Atomic Energy Agency. While negotiations would be more complex than those for nuclear arms, the stakes are equally high. With the right leadership, humanity could prevent an AI-driven catastrophe, just as it has avoided a thermonuclear war.