from the to-claim-copyright-or-not-to-claim-copyright dept
Andy Baio always digs up the absolute best stories. His latest involves layers upon layers of fascinating issues and legal questions. The key part, though, is that Jay-Z and his company Roc Nation, were able to convince YouTube to remove two “audio deepfakes” by claiming both copyright infringement and “unlawfully using AI to impersonate our client’s voice.” Both of these are highly questionable claims. But let’s take a few steps back first.
We’ve discussed how there seems to be a bit of a moral panic around deepfakes, with the idea being that more and more advanced technology can be used to create faked video and audio that looks or sounds real — and that might be used to dupe people. So far, there’s little evidence of the technology ever actually being used to really deceive people, and there’s plenty of reason to believe that society can adjust and adapt to any eventual attempts at using deepfakes to deceive.
Still, in part because of the media and politicians freaking out about the whole idea, a number of social media platforms have put in place fairly aggressive content moderation policies regarding deepfakes, so as to (hopefully) avoid the inevitable big media “expose” about how they’re enabling nefarious activities by not pulling such faked videos down. But, as we’ve noted in some of those previous articles, the vast majority of deepfake content these days is purely used for entertainment/amusement purposes — not for nefarious reasons.
And that’s absolutely the case with the anonymous user Vocal Synthesis, who has been playing around with a variety of fun audio deepfakes — just using AI to synthesize the voice of various famous people saying things they wouldn’t normally say (or singing things they wouldn’t normally sing). The creator releases them as videos, but it’s just a static image, and even when they’re “singing” songs, it’s without any of the music — just the voice. So, here’s Bob Dylan singing Britney Spears’ “… Baby One More Time”:
Some other people have taken some of those audio deepfakes and put them to music, which is also fun. Here are six former President’s singing N.W.A.’s “Fuck the Police”:
A few of the audio deepfakes use Jay-Z’s distinctive voice — and apparently Jay-Z or his lawyers got upset about this and issued takedown notices to YouTube on two of them. As I type this, those two videos (one of Jay-Z reciting the famed “To Be, Or Not To Be” soliloquy from Hamlet and another of him doing Billy Joel’s “We Didn’t Start the Fire”) are back up with YouTube saying that the original takedown notices were “incomplete” and therefor the video had been reinstated. But they were taken down originally, and it’s possible that more “complete” takedowns will be sent, so for the time being (as Andy Baio did) I’ll also point to the same content hosted by LBRY, a decentralized file storage system:
And here’s where things get odd. As Andy notes in his post (which is so detailed and worth reading), the takedown from Roc Nation made two separate claims: first that the videos infringe on Jay-Z’s copyright, and the second that each video “unlawfully uses an AI to impersonate our client?s voice.” But what law is being broken here? And if it was illegal to impersonate someone, a bunch of impressionists would be in jail. Andy goes through a detailed fair use analysis on the copyright question:
There?s a strong case for transformation with the Vocal Synthesis videos. None of the original work is used in any recognizable form?it?s not sampled in a traditional way, using an undisclosed set of vocal samples, stripped from their instrumentals and context, to generate an amalgam of the speaker.
And in most cases, it?s clearly designed as parody with an intent to entertain, not deceive. Making politicians rap, philosophers sing pop songs, or rappers recite Shakespeare pokes fun at those public personas in specific ways.
Vocal Synthesis is an anonymous and non-commercial project, not monetizing the channel with advertising and no clear financial benefit to the creator, and the impact on the market value of Jay-Z?s discography is non-existent.
We have talked about Conde Nast previously using a copyright claim to take down a deepfake of Kim Kardashian that was highly questionable, but at least in that case it was clearly making use of an original video from Conde Nast. Here, it’s not even clear that Roc Nation can say what the registered copyright they have that is being used here.
Andy’s post also includes an interview with the (still anonymous) creator of these videos, and I suggest reading the whole thing, but here’s one short snippet that I found super interesting:
Mainly, I?m just making these videos for entertainment. Sometimes I just have an idea for a video that I really want to exist, and I know that if I don?t make it myself, no one else will.
On the more serious side, the other reason I made the channel was because I wanted to show that synthetic media doesn?t have to be exclusively made for malicious/evil purposes, and I think there?s currently massive amounts of untapped potential in terms of fun/entertaining uses of the technology. I think the scariness of deepfakes and synthetic media is being overblown by the media, and I?m not at all convinced that the net impact will be negative, so I hoped that my channel could be a counterexample to that narrative.
As noted, YouTube has currently put the videos back up, but I imagine we’ll see a lot more of this in the near future.
Oh, and I forgot to mention that Vocal Synthesis originally announced the takedown by creating an audio deepfake of Barack Obama and Donald Trump explaining the situation.