MBW Views is a series of op-eds from eminent music industry people… with something to say.  The following op-ed comes from David Israelite, President & CEO of the National Music Publishers’ Association (NMPA). NMPA is the trade association representing American music publishers and their songwriting partners.

AI music companies are at a make-or-break moment, argues Israelite: come to the table with songwriters, as Udio and Klay have, or keep exploiting copyrighted work for free and face the growing weight of litigation and new legislation…

NMPA recently announced two groundbreaking industry-wide deals with AI music companies.

Both had important elements. The first, with Udio, values record label and publishing training data equally – a critical precedent. The second, with Klay, ensures the entire platform was licensed with publishers and songwriters before becoming available to the public. Another paradigm shift.

While this progress is significant, there still are many unlicensed AI platforms that pose a serious threat to songwriters and lawsuits and legislation will dictate the path forward. However, outside of these factors, these companies can determine that path for themselves.

It’s easy to conflate AI music companies as all good or all bad. The truth is that there is a broad range of functionality and postures towards the creative community.

Each AI business has an opportunity to license and grow alongside creators – enjoying their support and contributions – or to reject the very clear reality that the songs and data they require have value. As an industry must embrace those good actors who choose the former and come to the table.

When it comes to bad actors, none have been more brazen than Anthropic. The AI giant which trains its chatbot Claude on copyrighted lyrics has consistently held that this practice is fair use. Common sense clearly shows it is not. And the stakes for the company are enormous.

Rumblings of Anthropic’s upcoming IPO have predicted a potential trillion-dollar valuation. This unimaginable windfall is premised on the platform’s success in creating derivative work from existing copyrighted work.

The profit and effect on the market is staggering. As NMPA and others argued in an amicus filed in March, “AI-generated music already acts as a direct market substitute for human-created works, and that a functioning licensing market for AI training already exists — one that Anthropic has chosen not to participate in.”

Defenses of Anthropic’s fair use argument have been sophomoric and wrong-headed. In the Bartz v. Anthropic case, a judge equated scraping all of human creation to artificially create new versions to ‘teaching schoolchildren to write.’

Setting aside the fact that this compares the greatest songwriters and lyricists in history to providing a basic education, a more apt metaphor would be Anthropic using the world’s complete song cannon to teach students to read and write complex compositions, and take their teachers’ jobs. All while not giving their instructor a dime and using their new skills to collectively become a billionaire.

“AS AN INDUSTRY WE ARE AT A PRECIPICE. EVERYONE IS TAKING STOCK OF AI PLATFORMS’ POSITIONS TOWARD CREATORS AND DETERMINING WHAT THE NEXT DECADE WILL LOOK LIKE.”

DAVID ISRAELITE

Irony of ironies, Anthropic is currently suing Alibaba for stealing its capabilities. The AI giant knows and has argued that taking someone else’s design to create a competitor does not pass the commonsense test. The scale of distillation in particular is analogous.

“Distillation is a technique in which a less capable AI model is trained on the outputs of a stronger one to copy its behavior. … The Claude developer says operators affiliated with Alibaba and its Qwen AI lab generated more than 28.8 million exchanges with Claude through almost 25,000 fraudulent accounts.”

Call it distillation or scraping, models that train on proprietary material are not only stealing they are devaluing the work they’re benefitting from. AI platforms have made the very same argument in their own cases.

“OpenAI said in January 2025 that China’s DeepSeek may have used its data to train a rival model without permission. DeepSeek said it spent about USD $5.6 million training its model, against the more than $100 million OpenAI spent on GPT-4. Its debut wiped roughly $600 billion off chipmaker Nvidia’s market value in a single day.”

While lawsuits will direct much of the future, there is also serious legislation which could reign in the current wild west of AI.

The CLEAR Act – Copyright Labeling and Ethical AI Reporting – requires anyone using a training dataset to build a generative AI model to file a notice with the Register of Copyrights summarizing each copyrighted work used.

To give songwriters recourse when their works have been used and they were not notified, the TRAIN Act – Transparency and Responsibility for Artificial Intelligence Networks – creates an administrative subpoena process to let copyright holders obtain records to determine whether their works were used to train a generative AI model.

Finally, the COPIED Act – Content Origin Protection and Integrity from Edited and Deepfaked Media – directs NIST to develop standards for content provenance, watermarking, and synthetic-content detection and allows creators to prohibit training on AI without consent.

And in regard to deepfakes, NMPA is working alongside our broad industry coalition to push for the final passage of the No Fakes Act which recently was advanced by the Senate Judiciary Committee to be voted on by the full Senate. The bill creates a new federal property right for every American in their voice, image, and likeness lasting for life and up to 70 years after death.

There are additional bills that make important progress on legitimizing how AI platforms train and preventing them from stealing songwriter’s ideas and identities. However, the fastest way to make good with creators is to recognize the inevitable end of these legal disputes and legislative efforts and dispel the notion that there’s anything fair about using copyrighted work for free.

As an industry we are at a precipice. Everyone is taking stock of AI platforms’ positions toward creators and determining what the next decade will look like.

Litigating against bad AI actors and licensing good AI partners is not in conflict. NMPA will do both. Whether companies meaningfully engage songwriters will determine which are successful or pursue a race to the bottom. Either way, we will uphold the integrity and value of human created songs which have never been more essential.

Read on Music Business Worldwide.