Amicus brief urges US court to treat AI training inputs as a separate licensing market

What happened

Teodor J. Holmberg filed an amicus curiae brief in the consolidated copyright litigation against OpenAI and Microsoft in the U.S. District Court for the Southern District of New York (In re OpenAI, Inc. Copyright Infringement Litigation, before Judge Sidney H. Stein), urging the court to recognise the use of copyrighted works as AI training inputs as a separate licensing market.

The two arguments

First, Holmberg argues that preparing copyrighted works for AI training — including formatting, modification and tokenisation — could amount to creating derivative works under Section 106(2) of the U.S. Copyright Act. He contends the licensing market for works used as AI inputs must be assessed separately from the market for AI-generated outputs, and that converting text into tokenised data does not make the use transformative for fair-use purposes because it still relies on the original expressive content.

Second, in the alternative, he argues that if AI-input use is not covered by existing exclusive rights, the court should recognise it as a distinct property interest capable of supporting state-law claims such as unfair competition, misappropriation and unjust enrichment, without those claims being pre-empted by federal copyright law. He asks the court to permit further briefing on those claims.

Context

The brief points to licensing agreements by OpenAI and Microsoft with media companies, and arrangements involving Amazon, Google, Meta, Perplexity, ProRata.ai and Mistral AI, as evidence of an emerging commercial market for copyrighted material used in AI development. It relies on the Third Circuit's 29 September 2026 ruling in Thomson Reuters v. ROSS Intelligence, which found unauthorised use of headnotes as AI training data harmed a potential derivative market, and argues the 2025 Bartz v. Anthropic and Kadrey v. Meta decisions wrongly excluded the training-inputs market from the fair-use analysis. The filing comes amid competing summary-judgment motions by authors and news organisations on one side and OpenAI and Microsoft on the other.

What is not established

An amicus brief is an argument by a non-party, not a court finding. No ruling on the merits of the copyright claims has been made.