Content Filtering: Content filtering is the use of digital technology to act automatically on content that is shared, stored, or published by a digital platform or service provider. Nothing in the copyright law requires content filtering, and safe harbors in Section 512 of the DMCA ensure that internet service providers, including platforms for user generated content, are not held liable for their users’ activity so long as they respond appropriately to takedown notices. Nevertheless, some platforms use filters to enable private arrangements with copyright holders that include removing, demonetizing, or otherwise responding to content that matches a filter. 

Why it Matters: Content filters are a holy grail for large corporate copyright holders, who have asked congress to mandate filtering (and relatedly, to mandate the use of technology that would make filtering easier) in a wide variety of contexts. Examples include“broadcast flag” proposals that would have forced consumer technology companies to filter digital broadcasts and block recording of “flagged” content, to so-called “notice and staydown” proposals that require platforms not only to remove a particular copy of an allegedly infringing work in response to a notice from a copyright holder, but also to take steps to find and suppress all future copies posted to their systems. The spirit of filtering is alive and well in bills like NO FAKES (which has a staydown provision that public interest groups flagged as a threat to free speech) and California’s AB 412 (which requires AI developers to scan their training data using “digital fingerprints”). Filters promise automated copyright enforcement at scale. However, in practice content filters offer rightsholders much more than enforcement of their existing rights: it offers the chance to erase every pesky limitation and exception to copyright and replace the delicate balance of the law with total control. Filters can’t ask why a copy is being made or published or who (if anyone) will be the audience. They don’t know whether a film clip is being criticized or parodied, whether a song is mere soundtrack or the subject of a dissertation or a lecture. They typically can’t even tell if a work is used with permission. Copyright filters replace 300 pages of statute and 200 years of caselaw with a simple binary: is this a match or not? Under copyright law, that’s not the end of the analysis; it’s the beginning.