US Justice Department backs OpenAI fair use defense against New York Times

The US Justice Department siding with OpenAI and Microsoft on fair use is one of the most consequential legal signals yet for foundation models. If a federal court ultimately accepts this framing, it lowers the existential legal risk that training on unlicensed web text could be ruled wholesale infringement. That would solidify the de facto status quo: frontier labs scrape the open web, pay selective licensing fees where it is politically or commercially useful, and treat the rest as fair game.

For the AGI race, this favors the largest, best capitalized players. They already have the infrastructure and war chests to capitalize on permissive interpretations of fair use. A win for OpenAI here would make it harder for publishers to force broad licensing regimes that might have slowed model scaling or forced more synthetic and curated datasets. At the same time, it may provoke a backlash in other jurisdictions, particularly the EU, that could tighten text and data mining rules.

The longer term question is whether courts will treat all training equally or begin drawing lines around particularly sensitive corpora, like medical records or private communications. For now, though, the US government has made clear it views AI competitiveness as a national priority and is willing to bend longstanding copyright debates in that direction.