Four unions urge judge to reject blanket AI-training fair use in OpenAI case

Four entertainment unions have asked a federal court in New York for permission to intervene as friends of the court in OpenAI’s consolidated copyright litigation, arguing that AI training cannot be declared fair use as a blanket rule.

SAG-AFTRA, the American Federation of Musicians, Writers Guild of America West and Writers Guild of America East filed their proposed amicus brief on Oct. 29 in In re OpenAI, Inc., Copyright Infringement Litigation, before Judge Sidney H. Stein in the Southern District of New York. The court must approve the request before the brief becomes part of the record.

The unions collectively represent more than 250,000 media and entertainment workers. Their filing backs the plaintiffs, arguing that members’ work has been used to train AI systems that can copy and compete with it, and that the effects of unlicensed training reach beyond companies such as the News Plaintiffs.

Voice performers are central to the unions’ argument. The brief says actors are already losing bookings to synthetic voices, including reports of performers being hired and later replaced by AI in vertical micro-dramas.

It cites the National Association of Voice Actors’ 2026 survey, which received 1,379 responses. Twenty-one percent of respondents said they had knowingly lost a job to a synthetic voice. NAVA collected no identifying information from participants.

The responses also showed a more qualified attitude toward the technology. Thirteen percent said they had willingly allowed a third party to create a synthetic version of their voice, while 53% said they would consider an ethically created synthetic voice in future. The proposed brief said 94% had not made a synthetic version of their voice.

The unions want fair use assessed case by case. They contend that training models on human-created works in order to generate competing material weighs against fair use, and that a categorical ruling would suppress an emerging market for licensing training material.

Their argument also draws on bargaining provisions negotiated after the 2023 SAG-AFTRA and WGA strikes, when AI was a central issue. The unions say the settlements established minimum compensation for digital replicas and protections concerning fully synthetic characters. SAG-AFTRA and the writers guilds’ 2026 agreements include notice-and-bargaining rights for third-party licences of covered work used to train public-facing commercial generative AI systems.

The filing argues that a broad fair-use ruling would undermine those negotiated terms by treating training as a licensable use in which workers have a stake. It says OpenAI and Microsoft can afford licences, citing estimated valuations of about $1.4 trillion for OpenAI and more than $2.7 trillion for Microsoft.

The underlying consolidated case includes claims by The New York Times, other newspapers, The Intercept and authors including Ta-Nehisi Coates, Michael Connelly and Jodi Picoult. The unions are not plaintiffs, but say the result will affect the copyrighted works on which their members depend.

OpenAI has separately argued that fair use supports innovation and that AI training uses copyrighted works for a fundamentally different purpose without giving users substitute access to those works. The U.S. Copyright Office has said that licensing is one possible response to questions of consent and compensation in AI training, while noting that courts and policymakers are still at an early stage in addressing rapidly evolving technology and markets.