Here’s what’s actually verifiable: MLex’s headline asserts that a Chinese scholar is pushing back on the fair-use defense AI companies lean on in copyright fights. The reporting available for this piece is the headline alone — no name, no institution, no indication of whether the challenge is a law-review argument, a court filing, a regulatory submission, or testimony aimed at a specific jurisdiction. That’s a meaningful gap: “fair use” as a doctrine is a US-specific construct, and China’s copyright law has no direct equivalent, so a Chinese scholar’s intervention could be arguing something quite different — a critique of importing the fair-use frame globally, a defense of Chinese AI firms’ own training practices, or a challenge aimed at a specific pending case.
What’s independently established is the terrain this claim would land on. Hachette, Elsevier, Cengage and author Scott Turow sued Google in a New York federal court over Gemini training data, in a complaint Al Jazeera reported on July 15, 2026, that cites internal Google documents allegedly warning of up to $100 billion in potential fines. That suit followed a February 2026 attempt by Hachette and Cengage to join an existing 2023 authors’ class action. Meanwhile the doctrine’s track record in court is mixed: a federal judge ruled in 2025 that Meta’s use of the Kadrey plaintiffs’ books met fair-use requirements, while a judge denied OpenAI’s motion to dismiss the George R.R. Martin/Authors Guild suit in October, per Al Jazeera’s reporting.
A scholarly attack on fair use only matters if it names a mechanism courts haven’t already weighed — right now we have neither the scholar nor the mechanism.
What would validate MLex’s framing: the scholar’s name and paper, which jurisdiction’s doctrine they’re actually attacking, and whether any litigant — in the Google, OpenAI, or Meta suits — cites the argument. Absent that, this is a headline pointing at a debate, not yet evidence of one.
Chinese scholar challenges AI industry's fair-use defense in copyright debate
— MLex
