S. 51, although well-intentioned, would impose unconstitutional restrictions on free speech, create unworkable regulatory burdens and ultimately harm the very children it seeks to protect.
The bill’s vague definitions of “harm” and requirements that platforms modify their algorithms to reduce undefined risks amount to government censorship of protected speech. Furthermore, the monthly algorithm audits essentially require platforms to submit their editorial decisions for government approval. As in NetChoice v. Paxton and reaffirmed in Moody v. NetChoice, the Supreme Court has made clear that editorial algorithms constitute protected speech under the First Amendment.
Requiring monthly algorithm audits of complex machine learning systems would also impose enormous costs while providing little meaningful insight into child safety. The definition of “likely to be accessed by children” is so broad that virtually every general-audience website would qualify, from news sites to e-commerce platforms. This regulatory overreach would stifle innovation and impose crushing compliance costs on businesses of all sizes.