“Sam Altman advocates for a regulatory framework focused exclusively on safety testing for "extremely superhuman" frontier AI models, rather than broad regulation of less capable models.”