The First AI Copyright Appeal Is Heading to Court in June

For the past year, every major AI lab has been operating under legal uncertainty on one basic question: what does it mean to train a model on copyrighted material? A US district court gave the first answer in early 2025. In June, an appeals court will take another pass at it.

The Third Circuit has tentatively set June 11 for oral arguments in Thomson Reuters v. Ross Intelligence, making it the first US appellate ruling on whether using copyrighted works to train an AI model qualifies as fair use. The stakes for both the AI industry and rights holders are significant.

The case began when Ross Intelligence, an AI-driven legal research company, was found to have used Westlaw's proprietary legal headnotes to train its model. Ross had first sought a license; Thomson Reuters refused, since Ross was a direct competitor. Ross then worked through a third party, LegalEase, whose lawyers used Westlaw headnotes as a basis for creating AI training documents without copying them verbatim.

In February 2025, Judge Stephanos Bibas ruled against Ross at the district level. The court found that using copyrighted material in AI training was not transformative under fair use doctrine, and therefore constituted infringement. It was the first US ruling of its kind. Ross appealed, and the Third Circuit granted review in June 2025, becoming the first appellate court to take up an AI copyright case.

The appeal now turns on two core questions, as Norton Rose Fulbright's 2026 litigation tracker outlines: whether Westlaw's headnotes are copyrightable at all (they summarise judicial holdings, which are public documents), and whether internal use of those headnotes as training data might qualify as transformative use despite the district court's finding.

A ruling that reverses the district court would open training pipelines considerably, potentially allowing AI developers to use proprietary databases without licensing agreements, provided training is classified as transformative. A ruling that affirms it could force widespread changes to how AI companies source training data and accelerate the ongoing wave of licensing negotiations between publishers and labs.

Either outcome is likely to be appealed further. The Supreme Court already declined to revisit the human authorship question in Thaler v. Perlmutter earlier this year. AI copyright, it appears, will be settled one case at a time.