The A.I.-like-likeness-scanner (previously available to creators, politicians) is available for all celebrities to have takedown requests made to synthetic doppelgangers of them, regardless of whether or not they have a YouTube account.
YouTube activated yesterday morning what may be the most forceful action yet that the entertainment world has done to combat AI-generated celebrity content. The platform's facial recognition technology (which is essentially an additional version of Content ID) is being rolled out to entertainers, sports figures, artists and other agents representing them in Hollywood.
What YouTube is doing is turning policy language (that was related to using synthetic media deceptively) into a concrete searchable request process, so that celebrities can upload images of their own faces and search for synthetic versions of themselves created with AI. YouTube stated in its release announcing the availability of the tool, "This works similarly to Content ID", the platforms decade old copyright detection system that changed the way media companies monitored their content on-line. Rather than detecting audio prints or video clips, this tool detects faces (specifically faces that are believed to be artificially created or altered).
There seems to be some deliberateness to the timing. It has been eighteen months since we saw the first viral instances of AI-created impersonators of Tom Cruise. Six months ago California enacted legislation regarding digital replicas. And, there have already been many examples of fake movie trailer footage featuring unauthorized uses of celebrity likenesses racking up millions of views before people realized those were also synthetic. What YouTube appears to be doing is establishing itself ahead of regulatory actions by creating what Digital Watch Observatory refers to as "broader identity and rights governance" frameworks.
You do not need a YouTube channel in order to utilize this new tool. As reported by Tech Times , talent agencies and management companies can now submit claims on behalf of their clients as part of the new system, thus creating a whole new level of digital rights management that will extend far beyond YouTube's current creator-centric ecosystem. Similarly to when record labels received access to Content ID for direct uploading purposes, suddenly the enforcement mechanisms provided by the platform become a type of industry-wide infrastructure.
Get the latest model rankings, product launches, and evaluation insights delivered to your inbox.
It should be noted that there are specific limitations in place for this tool. According to CNET : "Not every flagged video will be automatically removed". YouTube will continue to perform manual reviews of claims for removal; which suggests that they recognize the complexity involved in making determinations of what constitutes harmful impersonation, versus what would qualify as legitimate parody or commentary. There isn't information available at this time as to what specific criteria YouTube will employ to determine which types of impersonation will result in removals. However, as Social Media Today points out users can "request the removal of AI-generated fake images and videos" via the system -- although such language does not promise removal. No Film School states that the purpose of the tool is focused on removing "fake trailers and unauthorized likenesses"; which directly references the major problem faced by studios as AI-generated movie preview trailers are spreading faster than marketing teams can remove takedowns for each one. Typically these fake trailers show deceased actors in unrealistic sequels, or living actors in movies that they would never consent to participate in; and create an intersection between fan-fiction and identity theft.
The mechanism behind utilizing this tool mirrors Content ID's method: upload your reference material; allow the algorithm to scan it; review potential matches; then submit takedown claims. In contrast to Content ID; which exists to protect economic interests; this system enters into less defined areas including personality rights and identity protection. Crescitaly Blog describes this as YouTube evolving from "broad policy language to a more searchable, claim driven process for protecting identities from deceptive AI usage"."
YouTube has yet to confirm whether the device is successful at removing artificial intelligence (AI) generated fake videos. At present, video creators use detection methods based upon AI-generated video flaws. Unfortunately, newer AI devices are able to remove these errors in their own production. Additionally, the company has not specified whether this system could recognize and remove other forms of artificially produced content such as facial substitution for another person, and/or entirely synthesized video and photos.
Additionally, by comparison to the rate at which they adopted Content ID, the Entertainment Industry's adoption of this technology seems rapid. Although YouTube mentions in their blog post about this tool, that it was "built with collaboration with the top agencies representing major talent" (and does not identify those agencies), the fact that they mention working with major agencies implies that major agencies were testing the technology prior to launching publicly; and presumably defining what type of artificial images and videos would be removed as part of an agreement between the agency and YouTube.
Major Talent Agencies now provide enforcement support for celebrities who wish to remove AI generated videos of themselves. In essence, Talent Agencies will serve as additional enforcement tools similar to record labels serving as law enforcement tools for copyrighted material. All takedowns must still undergo human review before being removed. The System identifies and removes both face swap artifical images/videos and completely synthesized video/images. To date there is no confirmation regarding whether this system will successfully detect and remove more complex AI generated images/video.