Authentic Brands Group, which holds Elvis Presley's music catalog, has filed suit against AI company Anthropic. The claim: the company used 274 Presley songs, including "Can't Help Falling in Love," without permission to train its models. The case, first reported by Billboard, was filed in court this past Friday.

What happened?

The suit was brought by the licensing organization that owns both Presley's music and his image rights. The charge is clear: Anthropic infringed the rights to these 274 works. The company's defense has not yet been made public, but AI firms are known to typically lean on a "fair use" argument in cases like this.

Why does it matter?

Anthropic positions itself as a "safety-focused" and responsible player in the industry. Even so, the company has recently faced a string of copyright lawsuits. Music publishers previously went to court over song lyrics, and authors over their books. This shows that the source of training data is no longer just an ethical debate but has become a direct legal risk.

What sets the Elvis case apart from the earlier ones is where the front line has shifted. Previous music suits mostly focused on lyrics and composition rights. What's at stake here is the entirety of registered works themselves and the clear ownership of the catalog. In text-based data, the "fair use" defense can be interpreted relatively flexibly, because ownership of news, blog, and book content is diffuse, and a claim of transformative use is easier to construct. Music, however, is different ground: the rights are centralized, licensable, and carry a defined monetary value. It is clear who owns a work and what it is worth on the market. That makes the "fair use" argument far more fragile when it comes to music.

The heart of the matter is this: when courts weigh fair use, they look at whether the use harms the market for the original work. For music, which already has a licensable, active market, this criterion tends to cut against the defendant. Because if a company used for free content it could have licensed for money, the "harm to the market" claim becomes concrete.

Who does it affect, and how?

This case concerns not just Anthropic but every major model developer, because the same training practice is widespread across the industry. The outcome could redraw the balance between data collection and licensing.

PartyLikely impact
AI companiesLicensing costs for training data rise; the "collect first, defend later" model grows riskier
Rights holdersMusic catalogs become a new revenue stream and a lever in licensing negotiations
Small developersUnable to afford licensing deals, they fall at a disadvantage against the big players
UsersLimited visible impact in the short term; may feed into product costs over the long term

The music industry holds a strong hand here. The rights are consolidated in one place, and the infringement claim is tied to a concrete number of works (274). That clarity makes the suit tougher than the diffuse text cases. While it's hard to prove the individual value of a publisher's thousands of articles, the value of a registered Elvis catalog is already fixed in the market.

What's next?

There are a few axes to watch. First, whether Anthropic will build its defense on fair use or head toward a settlement. AI companies have recently begun signing licensing deals with some rights holders; this case could accelerate that trend. Second, how the court will draw the line on fair use in music — the precedent that emerges will shape the entire industry's data strategy.

A cautious reading: while the visible aim of these suits is damages, their real function is to set a price. Rights holders are turning the courtroom into a negotiating table. The likely result appears to be AI companies increasingly turning to licensing data rather than scraping it. That points to a new order that closes the era of "free internet data" — costlier, but on firmer legal footing.