To Sue, Or Not To Sue — That Is The Jay-Z Deepfake Question

Jay Z may regret his action last month to silence a synthetic version of himself on YouTube reciting the “To Be, Or Not To Be” soliloquy from Hamlet.  Perchance the rap royal liketh not being ghosted by Billy Thee Shakespeare.  

At the end of April, Jay Z’s legal team haplessly tried to yank the gag off YouTube with a DMCA takedown notice, noticeably missing any legal basis under the US Copyright Act. You can’t copyright the mannerisms of someone’s voice.

Trademark or name and likeness rights might have provided better sound and fury, but you have to go to court to vindicate those rights. The beauty of a DMCA takedown procedure is how quick, quiet, intimidating, and effective it can be… if it works.

But it didn’t work. The video bounced right back up a day or two later as YouTube reportedly deemed the DMCA demands “incomplete,” a Google
GOOGL
spokesperson told The Verge. A spokesman for Jay Z’s Roc Nation and legal counsel have not responded to requests for comment.

Jay Z’s legal team also swung their DMCA billy clubs at a Billy Joel deepfake on the same YouTube channel, with Jay Z reciting “We Didn’t Start The Fire.” Maybe Billy Joel rubbed Jay Z the wrong way, but the Bard?

Where do you draw the line with deepfake voices?

At computers, it seems.  When a flesh-and-blood impressionist like Rich Little or Frank Caliendo nails a celebrity or politician, it’s a belly laugh. But when a silicon chip impressionist kills it, it causes conniptions. Weird Al, yes. Weird  Algorithms, not so much.

Jay Z may be rightly uptight, when you consider the possibilities and the realities.   

For instance, let’s say it’s a deepfake voice of a CEO calling an underling to order a transfer of hundreds of thousands of euros to a certain bank account. That happened last September. The duped executive “recognized his boss’ slight German accent and the melody of his voice on the phone,” and got scammed, said fraud expert Rüdiger Kirsch, in a Wall Street Journal report.

Audio deepfakes can be as evil and embarrassing as video deepfakes.  In 2017, celebrity faces superimposed on porn star bodies started showing up on Reddit and elsewhere on the Internet. It got the world’s attention. And now it’s got some in an already deeply worried world worrying about viral phishing scams that can jump species from email to voice with a highly efficient transmission rate thanks to convincing con-bots.

The Voice Synthesis channel on YouTube that posted the Elizabethan Jay Z jest, has other outrageous and funny (depending on your sense of humor) voice parodies, like Bob Dylan voicing Britney Spears’ “Baby One More Time” and five U.S. presidents rapping NWA’s “F**k tha Police”. Dylan wisely ignored the mockery, or perhaps chuckled. The same way he ignored or chuckled over many mock-ups of his mannerisms back in the day, like Stealers Wheel’s “Stuck in the Middle with You,” where Gerry Rafferty might have sounded slightly like a Dylan-bot invoking Rafferty’s “clowns to the left of me, jokers to the right” if a Zimmerman algorithm existed in 1973 like it does today.

Likewise, the five U.S. presidents or their estates have better things to do than fire off DMCA takedown notices at YouTube. Maybe some chuckled.

Maybe some knew about “Fair Use”, a concept under both U.S. copyright law and trademark law, each with its own rules, but both with a parody exemption from liability for infringement.  In other words, if you’re poking fun at the butt of the joke using their own words, music, or appearance, you’re probably inoculated against most lawsuits.  

But parody may pass into perfidy if the enterprise is more commercial than comedic. “Sound-alikes” on TV commercials come to mind. More in the realm of trademark and unfair competition law than copyright law, sound-alike cases have involved Tom Waits suing Frito Lay for having someone imitate his gravely voice for a TV commercial. That was after Tom turned down Frito Lay who had reached out to him. Clearly that was an infringement. They wanted to make money off of Tom’s famous vocal style and arguably confuse the public to think it was him.  With deepfakes, a new Frito Bandito comes to town, this time with heavier AI artillery. But the same laws should apply: if it’s more of a commercial rip than a comical ribbing, it’s more likely an infringement, whether the sound-alike comes from a human or software.

Musicians should be as concerned as rappers.

A music emulator developed by David Cope, a computer scientist, composer, and author at UC Santa Cruz, can spool out miles of convincing music in the style of whoever its programmer wants: from Bach chorale to Mozart sonata to Chopin mazurka, Joplin Rag, and even a work in the style of its creator, Cope. Any musician with a distinctive style is fair game.

Thankfully for musicians, computer-generated music is still not perfect and sometimes laughable. For instance, a jazz-bot developed by Sony in France emulates the late pianist Bill Evans with mixed results. Evans’ heavenly flights of pianistic virtuosity, often done while doped up on heroin, helped define Cool Jazz on Miles Davis’s “Kind of Blue” album.  Sony’s Evans-bot sounds more like it’s doped up on a cocktail of Thorazine and Windows 8. The lush chordings and rush of arpeggios are trademark Evans, but the ham-fisted dynamics and pointless melodies reveal no one human is home.

Unfortunately, the robots are getting better fast. The founder of the Voice Synthesis channel told Pitchfork that a new and improved version of the OpenAI’s jukebox software could imitate voices of anyone alive or dead as well as unique combinations of familiar voices. He said he anticipates more takedown requests.

Hamlet was written 400 years ago. Assuming civilization is still around 400 years hence, Jay Z’s artistry, his clever play-on-words and compelling verbal imagery, may still be on display in the world, while Billy thee Shakespeare may be long forgotten, depending on the direction of civilization and education, but one hopes they’ll still get a good parody.