The sandwich-shop analogy is rhetorically sharp and hard to refute on its own terms — society does demand safety proof for far smaller risks. But it papers over the hard part: unlike a deli, a frontier model's failure modes are largely speculative, and regulators still can't agree on what to test for. Worth noting that 'zero regulation' is US-specific — the EU AI Act already imposes obligations, while American state-level safety bills keep dying under industry lobbying. So the real fight isn't whether to regulate but how, ideally before a genuine incident forces rushed, poorly designed rules. Sanders' consistency keeps pressure on Washington, though with this Congress, sweeping federal AI legislation remains a long shot.