Hollywood groups challenge AI licensing defense in Times copyright case

A common licensing argument, different legal requests

On October 30, the Human Consent Foundation, a nonprofit co-founded by actor Cate Blanchett, joined talent agency Creative Artists Agency and musicians’ advocacy group the Music Artists Coalition in asking permission to submit a friend-of-the-court brief on licensing at scale. Blanchett, who has worked in film and theater for more than three decades, points to the contracts and permissions governing use of her image and voice as experience behind the coalition’s argument, described in a report quoting the brief.

The foundation also has a stake in the infrastructure it advocates. Its proposed Human Consent Standard would let rightsholders publish machine-readable restrictions, licensing terms and paths to permission for creative works, identities, characters and marks. The August 8 draft specification lists Blanchett among its editors.

That proposal has important limits. The specification calls itself a work in progress and expressly prohibits production implementation. Registry authorization, identifier administration and its licensing protocol have operational dependencies still to be finalized. It also requires owners and representatives to resolve ownership shares, territorial rights and approvals before publishing a consistent rights position. A common data format would not itself settle those underlying rights questions.

The group’s filing says it takes “no position on the ultimate disposition” of the pending motions. Its contribution concerns the feasibility of identifying and licensing rights. That is narrower than an unconditional demand that the Times win.

The separate proposed brief from nine music organizations, including industry associations the Recording Industry Association of America and National Music Publishers’ Association and performing-rights organizations ASCAP and BMI, explicitly asks Stein to find that the defendants’ training on the plaintiffs’ works is not fair use. The groups say their members own or control most music consumed in the United States. They are not parties seeking damages in this case, but their industry is pursuing AI companies including music generators Suno and Udio and AI developer Anthropic.

The Music Artists Coalition is part of the foundation-led filing, rather than the RIAA-led brief. Its involvement follows earlier advocacy: it backed music publishers against Anthropic in March and signed a June letter opposing the automatic inclusion of artists in AI deals. That record highlights a further question within licensing itself—whose consent a deal actually secures.

Entertainment unions took their own position on October 29. SAG-AFTRA, the American Federation of Musicians and the Writers Guild of America’s East and West branches urged rejection of a categorical fair-use exemption for AI training. The unions described their filing as defending more than 250,000 entertainment and media workers. Journalism organizations also submitted supporting briefs, while talent agency WME argued that AI use of creative material should be licensed, according to a report on the wider intervention.

CISAC’s 2024 royalty collections were dominated by music. These figures measure existing rights administration, not AI-training licensing revenue or clearance costs. Source: CISAC.

Existing markets test the practicality defense

The foundation-led brief argues that collective licensing solves precisely the transaction-cost problem raised by dispersed ownership: centralized organizations can administer repertoires through blanket licenses and automated rights information, rather than negotiating every work individually.

There is measurable scale behind that comparison, with a substantial music bias. CISAC, the International Confederation of Societies of Authors and Composers, represents collective management organizations that administer creators’ rights. Its 2025 collections report, based on data from 228 societies in 111 countries, reported €13.97 billion in worldwide creator royalty collections for 2024. Music accounted for €12.59 billion, or 90%; audiovisual works, drama, literature and visual arts made up the remainder. These collections demonstrate the scale of existing rights administration, but do not measure AI training revenue or the cost of clearing training rights.

The music brief also points to AI agreements involving several competing services: deals between all three major music companies and KLAY; Udio agreements with Universal Music Group, Warner Music Group and Merlin; and Suno agreements with Warner, BMG and Believe. It also cites ElevenLabs agreements with Universal, Kobalt and Merlin. The groups argue that these transactions show a functioning licensing market. The reported examples do not establish comparable prices, identical permitted uses or the cost of clearing material for a general-purpose language model.

Market harm is the other fault line. The music organizations argue that training enables synthetic substitutes for human work, challenging the defendants’ description of it as learning to communicate and spread knowledge. The Times’ allegations likewise focus on competing products. Neither claim is a finding resolving this case.

OpenAI presents the opposite account. In its October 30 statement collecting excerpts from supporting briefs, the company highlighted authors, academics, librarians, accessibility advocates and technology groups defending fair use. In the excerpts OpenAI published, supporters argue that training extracts statistical patterns for a different purpose from distributing the original works. Authors Alliance argues that competitive pressure from noninfringing outputs in the same genre is distinct from substitution for protected expression. Library and accessibility organizations also argue that restrictions would affect people relying on AI systems developed by others.

The Justice Department has also supported treating unlicensed text training as fair use. That position does not necessarily extend to the outputs of AI queries, a distinction that keeps the act of training separate from what a product ultimately supplies to users.

What Stein still has to decide

Fair use permits some reuse of copyrighted material without permission. The pending dispute concerns whether that exception covers the defendants’ training uses; showing that licenses can be sold does not itself settle the answer.

The music brief invokes a September Third Circuit decision rejecting a fair-use defense in Thomson Reuters’ AI-training case against Ross Intelligence. That gives the groups a relevant precedent to argue from, while the Times litigation still requires its own ruling on the uses before Stein.

Stein has yet to rule on either side’s requests. The briefs sharpen two questions for that decision: whether training should be understood as a different use of the originals or as a route to competing substitutes, and what weight to give licensing markets that already operate alongside the defendants’ fair-use claims.