Regulating AI’s Rogue Capabilities: A Matter of National Security and Constitutional Principle

The debate over regulating AI has reached a fever pitch in Washington, with tech companies racing to develop ever-more powerful models while policymakers struggle to keep pace. At the heart of this debate is a question that goes beyond technical nuances: whether the US government should wield unilateral power over AI restrictions or if Congress should establish clear laws and guidelines.

On one side, advocates argue that America’s early adoption of advanced AI technology will make it safer and wealthier for generations to come. However, critics warn that growing risks associated with AI models – from cyberattacks to catastrophic harm – necessitate robust regulation before these technologies are unleashed on the world.

The Trump administration has been inconsistent in its approach, at times advocating for accelerated innovation and minimal regulation while asserting national-security concerns to justify early government access to new models. However, this ad hoc approach lacks clear rules or processes for determining which models get restricted and who gets access to them.

This lack of transparency creates a perfect storm for corruption, as AI vendors may be pressured to alter their models to advance the president’s ideological or political goals. Corporations that use AI face high stakes, with perverse incentives to stay in the good graces of the administration or cultivate favor with key officials.

Congress has been slow to act on AI regulation, introducing bills but failing to gain traction. The bipartisan legislation introduced by Representatives Ted Lieu and Nathaniel Moran would require AI companies to maintain the ability to shut down technology that could cause catastrophic harm – a welcome step, but its sufficiency is unclear.

The dangers of presidential control over AI restrictions are starkly illustrated by Dean W. Ball’s words: “You should not expect the most powerful people in the world using the most powerful technology ever conceived in a way that is inscrutable to the public to turn out well… and you should see that dynamic as fundamentally inconsistent with a democratic republic.”

Congress’s abdication of responsibility on this issue is shortsighted at best – and potentially catastrophic. While reasonable people disagree about the best regulatory solution, all should agree that limits on AI should be dictated by the rule of law, not the whims of the sitting president.

Private bodies like Ball has proposed to audit frontier labs and ensure adherence to safety plans, or multinational standards for AI as suggested by Demis Hassabis, are possible solutions. Alternatively, a public regulator could vet models before release and monitor them once they’re in the world.

Ultimately, whatever regulatory solution emerges must be guided by principles of transparency, accountability, and democratic governance. Anything less would be a recipe for disaster – and a betrayal of the Constitution’s promise to safeguard our freedoms from the concentration of power.

As the world hurtles towards an AI-powered future, we cannot afford to wait any longer for Congress to act. The fate of our national security and our democracy hangs in the balance.