The Last Mile: Platforms, Speech, and Remedies

A digital-replica right becomes meaningful only when a person can use it after an unauthorized face or voice has appeared online. This chapter follows the last mile from an anonymous uploader to an online service provider, then to notice, removal, speech review, and a possible remedy. The goal is to see why platforms are a practical chokepoint without assuming that every notice proves infringement or that every replica can be removed without constitutional cost.

The problem is structural. The person who created a fake may be anonymous or impossible to locate. The platform that hosts or distributes the file may be the only actor capable of stopping its continuing availability. But the same platform may also host a documentary, parody, song, political message, or criticism that uses a replica in a protected way. Enforcement therefore has to manage two kinds of error: leaving a harmful replica online and removing lawful expression.

The Platform Is the Practical Chokepoint

Start with a simple sequence. Someone creates or obtains a replica and uploads it to a service. The service stores, transmits, links to, or displays the material. The person depicted discovers it and sends a notice explaining that the replica is of them and was not authorized. The service decides whether to remove it, and the person may seek an injunction or damages if the harm continues. The direct uploader is still the first target, but the service controls the place where the public keeps encountering the material.

That control is why the Copyright Office treats secondary liability as a central design issue. Secondary liability does not mean that a platform automatically becomes responsible for every user’s conduct. Traditional principles generally ask whether an intermediary knew about the unlawful activity and materially contributed to it, profited from it while retaining the ability to supervise it, or took affirmative steps that encouraged the unlawful use. Those questions connect responsibility to knowledge, contribution, control, and purpose rather than to mere existence of a hosting service.

The difficulty is that an online service may process an enormous volume of content and may not know whether a particular face or voice is synthetic, whose identity it represents, or whether permission was granted. A service can be technically capable of removing a file while lacking the factual basis to decide whether the complaint is accurate. It may also face repeated uploads of the same material under altered file names or accounts. The platform is close to the practical lever, but proximity does not eliminate uncertainty.

The Office recommends a notice-and-takedown system paired with a safe harbor. A qualifying service would be protected from liability when it expeditiously removes an unauthorized replica after receiving actual knowledge or a sufficiently reliable notification. The safe harbor would make removal less legally risky, while the removal obligation would make the new right more useful against anonymous or unreachable uploaders. The report recommends learning from the DMCA’s experience without copying every part of it, and specifically rejects importing the DMCA’s “red flag” knowledge standard because of interpretive problems in copyright practice.