This month’s submission for Local 802’s A.I. series is by Ken Hatfield, a member of Local 802 since 1977.

We have dealt with the disruptive innovations of the digital paradigm shift for a quarter century, so why are so many so alarmed by what Artificial Intelligence portends? Because AI raises the stakes exponentially for all creatives with tech’s most disruptive technology to date!

The way we create and consume music has been irrevocably altered by tech’s disruptive innovations. I acknowledge some of those innovations have been beneficial, but few argue that the meager compensation we currently receive for our recorded work is even remotely fair.

Anticipating how AI could affect our future requires examining the impact of tech’s previous disruptive technologies on our industry, starting with one many use to access digital music: streaming.

The convenience of streaming makes it a very appealing way to access recorded music. But it has serious downsides for those creating the music being streamed. The audio formats used for streaming limit metadata (like names of ALL musicians on recorded tracks) to three fields of data. This matters, because without identification how will the audience know it was you that played that incredible solo as a sideman? The audio files are compressed making them smaller, so they use less bandwidth. While this makes them cheaper to send via the airways for handheld devices, compression dramatically limits the dynamic range of the music.

The remuneration rates we receive for streaming are abysmal. Based on current rates set by a three-judge rate court, we (musicians/rights owners) are entitled to 0.0026 cents per stream for commercial subscription streams and 0.0021 cents for commercial non-subscription streams. Some streamers pay more, but even those rates are still in the thousandths of a cent range. On April 1, 2024, Daniel Ek (co-founder and CEO of Spotify) unilaterally pronounced that Spotify will not pay anything for tracks with less than one thousand annual streams, despite those “underperforming tracks” generating $40 million a year and the law requiring payments for all streams!

Streamers calculate royalties based on the number of plays, not the length of time your music is listened to. This discriminates against longer tracks.

Streaming platforms routinely divide royalties they distribute using a pro-rata formula, not a user-centric formula. This results in all royalties due being put in a big pie and divided by market share. That means the subscription fees paid by those that listen to artists with smaller market shares often go to artists these subscribers do not even listen to! This makes it impossible to financially support indie artists by accessing their music solely via streaming platforms.

In March 2024 Spotify began “bundling” music along with podcasts and audiobooks. By July 2024, Spotify had “reclassified” 98 percent of its subscription accounts as bundles. Other streamers quickly followed suit. This switch to bundling means subscription fees will be further divided, and musicians will receive even less than the paltry fees we currently get.

This raises two obvious questions: (1) is the price for the convenience of accessing music via streaming too high? and (2) who really pays that price?

Ever since Web 2.0 emerged and the Digital Millennium Copyright Act became law in 1998, digital service providers have used music as clickbait to data mine information collected by surveilling what we listen to.

Data mining and advertising are two primary revenue streams that have turned big tech into a trillion-dollar industry. According to the International Federation of the Phonographic Industry, and the Recording Industry Association of America, more people consume more music now than ever. Yet musicians make less today than ever.

Consider this: new release vinyl records currently sell for around $30 U.S. That may seem high compared to $10 a month for unlimited access to all the recorded music available on a streaming service. But let’s take a closer look. In 1975 vinyl records sold for $7.50. That 1975 $7.50 equals $44.05 in 2024 dollars. So, $30 is not a bad deal for the consumer who wants to own an analog physical music product with liner notes, photos, and full credits, etc. But compare that $44.05 to the  0.0026 cents musicians currently get from streaming, and you can see just how unfair and unsustainable it is. I need 11,538 subscription streams to generate what I get from the sale of one vinyl record, and I need 5,769 streams for what a single CD sale generates. A 2021 study by the British Parliament lamented “the pitiful returns musicians and songwriters receive from streaming.” That study concluded that 85 percent of all musicians with music on streaming platforms earn less than $250 a year from streaming.

Many compare streaming to terrestrial radio (which by the way never paid artists or labels in the U.S. In fact, they still only pay composers and publishers). But terrestrial radio encouraged folks to buy our records. It was not a replacement for buying recorded music as streaming often is. That fundamental difference disrupted the entire music ecosystem. Instead of a fair share of revenues going to rights owners and creators of the music, with streaming, the lion’s share of revenue generated by our work stays with middlemen (like streamers), middlemen that merely provide access to our work, middlemen that invest nothing in the creation of our work (unlike record companies that used to fund our recordings). Economists call this business model a “monopsony”.

In a “monopsony,” a middleman (like a distributor) is strategically placed to pressure both vendors (those making products) and consumers. That middleman does not invest anything in the actual product; they only invest in their distribution network. Yet that middleman receives the vast majority of any revenue the product generates!

Virtually all DSPs (including streamers) currently operate as monopsonies!

Tech’s history of exploitation is not limited to musicians wanting their music to be noticed online. Tech companies seduce us into accepting terms of use agreements they know (from a quarter century of surveillance) we don’t bother reading. They entice us by offering something they know we can’t resist: free stuff. But is it really free? As Jaron Lanier famously said, “If you’re not paying for the product, then you are the product”. How many of you with Gmail accounts ever read the user agreement? How many read the “fine print” before accepting either the “cookies” (which are essentially tracking bots) required to fully access many websites, or the apps we upload to our devices?

Tech’s history of exploitation is why so many are so alarmed by the rapid spread of AI and the existential threat it poses to what we do as professional musicians.

AI is not just a threat to our livelihoods. It’s an existential threat to our very identities as creative artists because it uses our work to train itself to produce acceptable replacements of our work, and it does not even credit us for our unique creative ideas which form the heart of such “cloned works”! When your unique ideas are spread far and wide throughout the culture by nameless AI-generated pablum, try convincing anyone they were originally your ideas. And then try getting paid for your ideas!

Given tech’s history of exploitation, can we trust them to equitably employ the most disruptive technology yet developed? If past performance is any indicator, one can only surmise that they cannot be trusted with such power if it is unbridled.

Here are some suggestions for protecting musicians from AI:

(1) The American Federation of Musicians should revise their recording contracts to include requirements that all recordings recorded under AFM contracts must employ watermarking technology to identify all participants on all union recordings. Watermarking (added during the mastering process) securely embeds any metadata you choose (like ISRC #s, rights owners, featured artists and sidemen) into the actual audio files that comprise a digital recording. Recordings predating watermarking can either be remastered (adding the watermarked data) or fingerprinted retroactively (via the same software Shazam uses to audibly identify recorded tracks). With fingerprinting, you upload your metadata for recorded tracks to various databases when registering those recordings. Without watermarking or fingerprinting, even if we get the legislative remedies proposed below, tech will keep the money due many musicians whose work is used to train AI, by claiming they can’t identify any but the featured artists. Remember the black box funds the streamers were holding onto because they supposedly could not identify the rights owners for recordings they were streaming? Resolving that led to the creation of the Mechanical Licensing Collective (MLC) established under the Music Modernization Act. The MLC now distributes those black box funds.

(2) For watermarking to be effective, Congress needs to pass laws requiring compensation for all participants on recordings used to train AI. These “reuses” will generate revenue, so those whose work trained AI should share in the wealth their work helps generate. The compensation for sampling could serve as a model for such payments. (See Local 802’s A.I. Resource Page for current legislative efforts.)

(3) Congress also needs to pass laws requiring disclaimers for all music “product” generated by AI. Such disclaimers notify the consumer when the music they are buying or streaming, etc., was generated by AI, not created by human beings. Again, there is a model for such disclaimers: the GMO stickers we see in grocery stores for food that contains genetically modified organisms. For physical product (CDs, vinyl, etc.) the disclaimers could be stickers. For audio only (streaming, broadcast, etc.) they should be audible announcements before and after a set of tracks are streamed, broadcast, or played, along with visual screen disclaimers.

(4) There are additional actions needed to prevent DSPs from circumventing the intentions of any proposed legislative AI protections. Congress needs to eliminate the safe harbors from two laws dating back to the dawn of internet commerce: section 512 of the Digital Millennium Copyright Act (DMCA) from 1998 and section 230 of the Communications Decency Act (CDA) from 1996. These safe harbors make it impossible to hold DSPs liable for illegal activities perpetrated by users of their platforms. Even when the DSPs facilitate, encourage, or tolerate illegal activities, they can’t be held liable, because the DSPs only allowed the offending posts on their platforms, they did not actually post them themselves. But they do profit from potentially criminal posts ranging from sex trafficking to copyright infringement. And that’s the problem: there’s no economic incentive to do the right thing. In fact, all the economic incentives support tolerating, encouraging, promoting, and monetizing such reprehensible behavior. The elimination of these safe harbors will engender widespread support even outside the music community because we all see how bad things have become online with misinformation and disinformation, and we all know it needs to be fixed. Such broad support will greatly aid our fight for justice in the digital domain.

We must educate our fellow musicians and the general public about these issues. We need to advocate for protections of our rights. We need to develop solidarity among all musicians, union and nonunion alike. Combined with the protections proposed above, these actions will prove fundamental to creating and maintaining a sustainable musical ecosystem, one where all participants share equitably in the revenues our creative work generates.

Ken Hatfield has been a member of Local 802 since 1977. Send feedback on Local 802’s A.I. series to Allegro@Local802afm.org.

OTHER ARTICLES IN THIS SERIES:

A DEEP DIVE INTO HOW A.I. AFFECTS MUSICIANS

“It all sounds the same”

“Artful” Intelligence

“How are you going to stop AI from stealing our jobs?”

Unleashing Creativity