Comment by Guvante

The law in question was very specific and not as broad as you imply here. In fact more specific than the Redittor implies too since the law doesn't cover audio models in general only those purpose built to replicate a particular individual.

As the law is specifically targeting models made to duplicate an individual it isn't hard to provide sufficient evidence to clear the hurdles required of restrictions on free speech as examples of the negative effects are well documented and "speech model for an individual" isn't a broad category.

Also I would point out first amendment isn't used by the US to protect software delivery anywhere. Instead it is Congress explicitly encouraging it through the laws it passes.