A typical safety argument is that frontier models are uniquely dangerous, open models are especially dangerous, and therefore only a handful of enormous companies should be allowed to build good models.
But then a competitor ships and yesterday’s forbidden model becomes suddenly available for everybody to use.
And that’s why open models are very important, even if they lag in perpetuity behind the best closed models. No frontier lab owns the frontier for long. It is important that policymakers do not turn a temporary technical lead into permanent regulatory capture.
If the current incumbents would rather lobby politicians than advance, an open ecosystem should advance without them.