AI Risk Regulation Should Start With Testable Causal Mechanisms

An argument in the debate over AI risk says calls for regulation or slowing progress should explain what could go wrong, why it would happen, and how. That would allow the proposed risks to be examined, challenged, and potentially quantified. Some former critics of regulatory capture are now calling for regulation in response to recent AI developments.

The argument separates current employment claims from broader catastrophic scenarios. It says available evidence does not show that AI is causing job loss and may instead point to job creation, while the possibility that AI could end the world remains much harder to falsify and is still uncertain.

The uncertainty is compared with the early development of nuclear weapons, when their effects were not fully known. The argument rejects a zero-risk approach to decision-making and says it is not obvious that the federal government would manage a severe AI problem better than private citizens or companies. These are opinions, not established conclusions.