When AI is subject to less regulation than your fridge, something is startlingly wrong | Richard Hermer

The article raises concerns that the current pace of AI development presents risks of an "altogether different magnitude" compared to traditional national security threats. It distinguishes between "existential risk," which involves AI enabling mass harm or determining humans are inefficient, and "existential challenge," where AI could erode human agency with profound consequences for economies and national security. The author states the full panorama of these risks is not yet known, partly because AI is rapidly developing to a level humans may not understand or control.

A central argument is that AI regulation lags significantly behind its development, contrasting it with the stringent safety standards for everyday consumer products. The piece highlights a perceived "limited time window" to address these issues, noting that public discourse on AI risks gained traction recently after incidents such as the Hugging Face hack and a key resignation from Anthropic. This moment is framed as a critical opportunity for increased urgency in international coordination.

The author advocates for robust international law and meaningful public engagement to shape AI's future, rather than leaving its direction to a few states or tech giants. The comparison to the 2008 financial crash, where human understanding of complex financial systems failed, underscores the potential for devastation on a much larger scale with AI. The piece concludes by stating that slowing down frontier AI is "blindingly obvious" until its risks can be clearly managed.