The “Blue Wave” Could Change the Rules for Music AI

A historic blue wave has broken across the United States. Is there an opportunity for musicians? Photo Credit by Michael Whalen

A skeptical but hopeful look at the opportunity to protect music and the people who make it.

The Democratic victory in the midterms gives Congress a chance to change the terms under which AI uses music. Over the next two years, lawmakers could put real protections around copyright, a performer’s identity, and the money generated from creative work.

Wall Street has already reacted. On November 5, the Nasdaq fell roughly 5.5% as investors worried about tougher AI regulation and restrictions on data-center expansion. The election helped cool an AI trade already facing questions about enormous spending and uncertain returns.

That doesn’t mean AI development has suddenly slowed. Markets can change direction quickly. But investors are starting to account for the possibility that this industry will face more scrutiny.

I’m hopeful that a more deliberate pace could give us time to get a handle on what AI is already doing and make better decisions about what comes next. Musicians need to understand where their work is going, what they are agreeing to, and whether they will be paid. Those are ordinary business questions. The technology shouldn’t make them impossible to answer.

There is an opening here, although the political limits are real. President Trump will still have a veto. Democrats will have disagreements of their own. Several existing proposals have bipartisan support, but sponsorship and passage are very different accomplishments.

The most useful agenda would focus on ten protections that work together, from the moment music enters a training dataset to the moment a musician needs to enforce a right.

It starts with clear rules for commercial training. Copyright already controls certain kinds of copying, subject to exceptions including fair use. How those rules apply to particular AI practices remains disputed. Calling a process “training” doesn’t settle the legal question, and neither does assuming every use of copyrighted material is automatically illegal.

One court decision shows how complicated this can get. In the 2025 Bartz v. Anthropic case, the judge found the training use at issue fair while rejecting a fair-use justification for acquiring and retaining pirated books in a separate library. Different parts of building the same system produced different answers. It was not a nationwide ruling that resolved music training.

I would support a permission requirement for defined commercial uses of protected music in generative-AI training, with carefully written exceptions. Congress would need to address both the composition and the recording, since permission concerning one does not necessarily cover the other.

Senator Marsha Blackburn’s broader AI discussion draft proposes excluding unauthorized AI training from fair use. That is a serious proposed change, and its scope needs close examination.

A workable rule would give copyright owners a decision before the use takes place. It would also give developers a clearer basis for budgeting and building their products. Right now, too much depends on who can afford to keep arguing in court.

That leads to access to training records. A right is difficult to enforce when the other side has all the paperwork.

Imagine hearing a commercial system produce music that sounds remarkably close to your catalog. That resemblance may not prove your recordings were used, much less that an infringement occurred. You need evidence, and the developer may be the only party holding it.

The CLEAR Act would require disclosures about covered copyrighted works in training datasets. The TRAIN Act would create a subpoena process allowing owners to obtain records concerning their own material. Neither would automatically establish infringement. They would help answer the basic question of what happened.

Congress should build on those proposals with record-retention requirements and independent audits. Sensitive information could remain confidential while being available to qualified investigators or a court.

The coverage also needs work. CLEAR generally limits its definition to registered works and certain older recordings. Plenty of music has copyright protection without registration. Those works should be included in a useful transparency system.

The law also needs to separate ownership of a recording from control over a performer’s identity.

A company may own the master recording. That does not fully answer whether it should be able to manufacture new performances using the singer’s voice. State protections vary, leaving performers with different options depending on the circumstances.

The NO FAKES Act would establish federal voice-and-likeness protections and a process for addressing unauthorized replicas. Its progress has also exposed disagreements about free speech. Reporting, criticism, parody, and other protected expression need room to operate.

The protection should reach the session singer whose voice has professional value even if most listeners don’t know their name. It should include a usable removal process, because a lawsuit that takes years offers limited immediate help when a fake performance is circulating today.

We also have to leave room for musical influence. Copyright generally does not give someone ownership of a style. Protecting a singer against unauthorized replication is different from giving an artist control over a genre, a production approach, or a familiar musical vocabulary.

Even a strong identity right could be weakened by the contract a musician signs. That makes specific, informed consent essential.

A session agreement should not quietly authorize unrelated synthetic performances for years to come. Uploading a recording through a distributor should not become a guessing game about what else the company can do with it.

California’s restrictions on certain digital-replica contract provisions and SAG-AFTRA’s sound-recording protections offer useful examples. Their reach, however, depends on the applicable law or agreement.

A national standard should require separate, understandable authorization covering the use, duration, compensation, and parties involved. It should explain what happens when permission expires. A materially different use should require a new agreement.

Musicians could still make AI deals. They would have a better chance of understanding the deal they were making.

Then comes a defined share of the licensing money. This deserves more attention than it usually gets.

A licensing agreement can satisfy the copyright owner while leaving the people who made the music with very little. Copyrights can be transferred, and qualifying work made for hire can belong to an employer from the beginning. A musician’s payment rights often depend on contracts written long before these uses were contemplated.

So a headline announcing that a catalog has been licensed to an AI company tells us almost nothing about the checks going to individual creators.

Congress should examine a payment right for defined categories of people whose work is covered by these agreements. Songwriters, recording artists, producers, and session musicians would need separate consideration. Their contributions and existing rights are different.

The accounting should cover advances, equity, and other benefits as well as ordinary licensing fees. Creators should receive statements and audit rights. Lawmakers would also need to decide whether the payment right could be waived or swallowed by contractual deductions.

This would be a substantial change. It would require a defensible way to allocate money, rather than a promise that every generated note can somehow be traced back to one recording. The useful goal is an accounting system people can understand and check.

That payment system would be stronger if independent creators had a practical way to negotiate together.

A songwriter or small label may have little influence over the terms offered by a large technology company. Voluntary collective licensing could help, provided the organizations involved answer to their members and make their fees and payment formulas clear.

The White House’s March 2026 legislative recommendations already discuss collective licensing arrangements and an antitrust exemption. There is an existing policy idea to work from.

Creators should understand how to join, what participation authorizes, and how to leave. A compulsory license would raise a different question because it can remove the owner’s ability to refuse a use.

The system should also be manageable for smaller developers trying to obtain permission. Otherwise, Congress could accidentally build a market in which only the biggest catalog owners and technology companies can afford to participate.

Musicians using AI in their own work need clear ownership rules as well.

The Copyright Office recognizes that qualifying human contributions can be protected when AI assists the process. Material without sufficient human authorship does not become copyrightable simply because someone supplied a prompt.

That distinction should be reflected in understandable registration and disclosure procedures. A composer should be able to identify the human-authored elements of a project without being told that any AI involvement disqualifies everything.

There should also be remedies for knowingly false ownership claims. Musicians could otherwise find themselves fighting automated claims over generated material that the claimant has no legal basis to control.

An output’s lack of copyright protection does not establish that it is safe to use. It could still reproduce protected music or misuse someone’s identity. Those are separate questions, even when a software company makes the transaction look simple.

The next protection is reliable attribution and disclosure throughout distribution.

Information about a recording’s origins, contributors, and authorized uses should survive its passage through distributors and platforms. Listeners should also receive useful information when a performance has been generated or materially altered.

The COPIED Act addresses provenance information — records of where content came from — and certain unauthorized uses of content carrying it. The AI Labeling Act would require disclosures for covered generated content. Both offer starting points, although COPIED’s dependence on qualifying provenance information raises concerns about older and unmarked catalogs.

The labels need to mean something. A synthetic lead singer and a human recording cleaned up with software should not receive the same vague warning. Creators also need a way to correct an inaccurate label.

And a disclosure is not a license. Identifying a cloned voice does not establish permission to use it.

Once music reaches a platform, accountability for streaming fraud and withheld royalties becomes just as important.

Existing law can already reach fraud involving AI music. Michael Smith’s March 2026 guilty plea in a scheme involving generated songs and automated streams demonstrates that.

What musicians need beyond prosecution is a clearer account of how platforms detect suspicious activity, withhold earnings, and handle recovered money. Those systems should be auditable, and consequential payment decisions should come with explanations.

There also needs to be a meaningful appeal. Being wrongly flagged can create an immediate financial problem for an independent artist. “The system detected something” is not enough information to resolve it.

Generated music and fabricated listening activity are different issues. Rules should target the behavior that diverts royalties while protecting legitimate releases from careless accusations.

All of this depends on affordable enforcement. It should be designed into the law from the start.

Congress needs to specify who can bring a claim, what evidence they can obtain, which remedies are available, and when legal expenses can be recovered. A new right with no practical route to enforcing it would leave musicians in a familiar position: legally protected on paper and largely on their own.

The Copyright Claims Board offers a smaller-claim process, but respondents can opt out. It does not provide a compulsory, inexpensive forum for major AI disputes.

The Hawley–Blumenthal AI Accountability and Personal Data Protection Act proposes an individual right to sue over specified unauthorized uses of covered data, including copyrighted creations. Its broad language needs careful coordination with existing copyright rules, but its attention to individual remedies is worth pursuing.

Public enforcement and direct claims by affected creators could work together. Appropriate damages and fee recovery could make legitimate cases viable. Safeguards against abusive claims would help prevent the same process from being used to intimidate smaller artists.

Getting these protections into place will take sustained work through 2027 and 2028.

Early 2027 should be used to revive unfinished bills, settle disputed definitions, and give independent musicians a real role in the drafting. Proposals that remain unenacted when the current Congress ends will need to be reintroduced.

I would give the first legislative attention to training disclosures, digital replicas, and specific consent. Those proposals address identifiable problems and already have substantial drafting behind them. Training liability, creator-payment rights, and collective negotiation would probably require more extensive bargaining.

Implementation needs to begin alongside that legislative work. Congress should account for enforcement funding, required regulations, technical standards, and complaint procedures before announcing that a problem has been solved. Some protections could be operating during 2028. Others could remain in rulemaking or litigation into 2029 and beyond.

Election Day, November 7, 2028, is a reasonable point to judge progress. By then, we should be able to see what passed, what actually took effect, and what remains blocked.

The blue wave gives Democrats more control over the agenda. Whether that produces useful law will depend on the deals they make, the details they defend, and their willingness to stay with the work after the announcement.

A more measured pace could help us make better decisions about AI’s present and future. By the next election, working musicians should be able to explain how the new rules changed a contract, protected a performance, or put money in their hands. That is the progress I’ll be looking for…